Last couple days I went through a round of porting newLISP to AIX. Is anyone interested in the changes, perhaps folding them into the release?
I had a couple comments about the code base with regards to porting, incase the developers are interested:
1. The use of #ifdefs for platforms is a bit overused. While I'm not a big supporter of autoconf'ing everything, its general model of #ifdef'ing the capability/requirement, rather than the platform seems to work better for portable projects. #ifdef _READLINE is okay, but
2. The usage of "opsys" in newlisp.c is restrictive. Since this ID is referenced all over the tests as stuff like
For the most part however, the code is portable ANSI C that, with a few minor changes, compiles under both gcc and the native IBM compilers under AIX4.3 .3. The qa-net test fails currently, I've not added debug code to see why as of yet. And I'll be building/testing under AIX5.X later this week.
I'd be happy to publish/discuss my changes if there is interest from the community. Thanks!
Hi csstup,
Yes the If then else is not my game either ;-) quite a puzzle when more
OSes join the club indeed...
The qa test you should first just leave as they are, if it compiles without big
errors qa tests can be done ;-)
But I think Mr. Guru himself will certainly give you a reply..
Norman.
Hi csstup
'opsys' is referenced in the C source a few times, mainly to feed into the newLISP 'sys-info' call in nl-filesys.c. Later in the code it gets incremented by the bits for UTF-8 and mode and id the versin is a library.
In newLISP .lsp source a built-in variable exists, called: ostype to code platform specific newLISP (mostly necessary for Win32).
I suggest you add:
#ifdef AIX
int opsys = 10;
char ostype[]="AIX";
#endif
in newlisp.c and create a makefile_aix with -DAIX used in it. Then create a
#ifdef AIX
...
#endif
section in newlisp.h and use #ifdef AIX and #ifndef AIX throughout the C source code.
The current system is not perfect but has served well through the years. I agree that #ifdef <capability> is much better then #ifdef <platform> through out the code, but for the makefiles it is better to have -DAIX, -DSOLAIRS, -DLINUX etc. The best probably would be to have #ifdef <platform> only be used in newlisp.h to bracket a group of #define <cabalities> for each platform.
In part this has already happened (i.e. for vasprintf(), nanosleep() etc). If a specific other capability comes to your mind which could lead to take out more of #ifdef <platforms> throughout the C code let me know.
You will see that most platform dependent issues are in nl-filesys.c.
If you send me a package compiling on AIX based on the latest development release, I can modify/integrate your changes for the next development release. Just "#ifdef AIX" you stuff into it and I see how I can streamline things further if you haven't done already. I will then send you a package back to make sure things work on both side.
Over the years platforms have come and gone, it would be nice to have AIX among the standard makefiles as it probably will still be around for some time, and have it detected too in the 'build' and 'configure' scripts. These script use the UNIX utility uname to detect the platform.
Last not least very welcome to newLISP and thanks for participating in porting efforts ;-)
Lutz
Quote from: "Lutz"
'opsys' is referenced in the C source a few times, mainly to feed into the newLISP 'sys-info' call in nl-filesys.c. Later in the code it gets incremented by the bits for UTF-8 and mode and id the versin is a library.
In newLISP .lsp source a built-in variable exists, called: ostype to code platform specific newLISP (mostly necessary for Win32).
I suggest you add:
#ifdef AIX
int opsys = 10;
char ostype[]="AIX";
#endif
I thought 10 would be the logical choice, but a bunch of the tests all compare for the sys-info return value(ie, opsys), which makes them fail in wierd ways when set to 10. Perhaps using something like <15 for all Unix, 16-31 for Windows, etc, to not interfer with the flag usage of the variable as well.
For now I'm reusing 4, since its semi close to Solaris, and the tests will all run. The tests were *key* to verifying the various porting issues I came across.
Quote
#ifdef AIX
...
#endif
section in newlisp.h and use #ifdef AIX and #ifndef AIX throughout the C source code.
Yep, already made a makefile_aix, changed makefile, build, etc. _AIX is defined by default when compiling under AIX (both GCC and native compilers) so I always use that.
Quote
Right, thats how I usually do it. One .h that list all the capabilities based on the -DSOLARIS define at compile time, which then all other source is based on those caps/requirements. Thats how autoconf's config.h works.
Quote
Will do. I'll get the prelim stuff packaged up shortly. Its built against 9.1.1, is that okay?
Quote
Already changed build to detect uname and use makefile_aix as required.
Quote
Looking forward to helping out where I can. Thanks for the reply!
Just FYI, build from xlc (native compiler):
./build
Discovered AIX
make -f makefile_aix
xlc -c -g -O2 newlisp.c
689 1500-010: (W) WARNING in main: Infinite loop. Program may not stop.
xlc -c -g -O2 nl-symbol.c
xlc -c -g -O2 nl-math.c
xlc -c -g -O2 nl-list.c
xlc -c -g -O2 nl-liststr.c
xlc -c -g -O2 nl-string.c
xlc -c -g -O2 nl-filesys.c
xlc -c -g -O2 nl-sock.c
xlc -c -g -O2 nl-import.c
xlc -c -g -O2 nl-xml.c
xlc -c -g -O2 nl-web.c
xlc -c -g -O2 nl-matrix.c
xlc -c -g -O2 nl-debug.c
xlc -c -g -O2 pcre.c
xlc newlisp.o nl-symbol.o nl-math.o nl-list.o nl-liststr.o nl-string.o nl-filesys.o nl-sock.o nl-import.o nl-xml.o nl-web.o nl-matrix.o nl-debug.o pcre.o -lm -ldl -o newlisp
And from gcc:
./build
Discovered AIX
make -f makefile_aix
gcc -Wall -pedantic -Wno-uninitialized -Wno-long-long -fno-strict-aliasing -c -g -O2 newlisp.c
gcc -Wall -pedantic -Wno-uninitialized -Wno-long-long -fno-strict-aliasing -c -g -O2 nl-symbol.c
gcc -Wall -pedantic -Wno-uninitialized -Wno-long-long -fno-strict-aliasing -c -g -O2 nl-math.c
gcc -Wall -pedantic -Wno-uninitialized -Wno-long-long -fno-strict-aliasing -c -g -O2 nl-list.c
gcc -Wall -pedantic -Wno-uninitialized -Wno-long-long -fno-strict-aliasing -c -g -O2 nl-liststr.c
gcc -Wall -pedantic -Wno-uninitialized -Wno-long-long -fno-strict-aliasing -c -g -O2 nl-string.c
gcc -Wall -pedantic -Wno-uninitialized -Wno-long-long -fno-strict-aliasing -c -g -O2 nl-filesys.c
gcc -Wall -pedantic -Wno-uninitialized -Wno-long-long -fno-strict-aliasing -c -g -O2 nl-sock.c
gcc -Wall -pedantic -Wno-uninitialized -Wno-long-long -fno-strict-aliasing -c -g -O2 nl-import.c
gcc -Wall -pedantic -Wno-uninitialized -Wno-long-long -fno-strict-aliasing -c -g -O2 nl-xml.c
gcc -Wall -pedantic -Wno-uninitialized -Wno-long-long -fno-strict-aliasing -c -g -O2 nl-web.c
gcc -Wall -pedantic -Wno-uninitialized -Wno-long-long -fno-strict-aliasing -c -g -O2 nl-matrix.c
gcc -Wall -pedantic -Wno-uninitialized -Wno-long-long -fno-strict-aliasing -c -g -O2 nl-debug.c
gcc -Wall -pedantic -Wno-uninitialized -Wno-long-long -fno-strict-aliasing -c -g -O2 pcre.c
gcc newlisp.o nl-symbol.o nl-math.o nl-list.o nl-liststr.o nl-string.o nl-filesys.o nl-sock.o nl-import.o nl-xml.o nl-web.o nl-matrix.o nl-debug.o pcre.o -lm -ldl -o newlisp
Both compile clean with no warnings after the changes.
Thanks Csstup, the compiles look nice. Either send me the package to lutz at nuevatec dot com as attachment or let me know in a private message on this board where I can retrieve it.
Lutz
After modifying newlisp.c and ... as described in this thread I have tried it but got some errors?
Is there some sort of howto compile or installation guide for AIX?
Can you dump the compile output here and the make file you used?
Gives a little more insight ;-)
makefile_aix_utf8
###########################################################################################
# makefile for newLISP v. 10.x.x on AIX woth UTF-8 support
#
#
OBJS = newlisp.o nl-symbol.o nl-math.o nl-list.o nl-liststr.o nl-string.o nl-filesys.o
nl-sock.o nl-import.o nl-xml.o nl-web.o nl-matrix.o nl-debug.o nl-utf8.o pcre.o
# use following for UTF-8 support and add nl-utf8.o to the OBJS line
CFLAGS = -Wall -pedantic -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX
CC = gcc
default: $(OBJS)
$(CC) $(OBJS) -lm -ldl -lrt -lnsl -o newlisp
.c.o:
$(CC) $(CFLAGS) $<
$(OBJS): primes.h protos.h makefile_aix_utf8
###########################################################################################
I have added
#########################
#ifdef AIX
int opsys = 4;
char ostype[]="AIX";
#endif
#########################
to 'newlisp.c' and
########################
#ifdef AIX
#define OSTYPE "AIX"
#endif
########################
to 'newlisp.h' and
#######################
aix:
gmake -f makefile_aix_utf8
#######################
to 'Makefile' and
######################
AIX)
echo Discovered AIX:
gmake aix
exit
;;
######################
to 'build'.
I got the following link error message:
#############################################################################################
.............
gcc -Wall -pedantic -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-import.c
gcc -Wall -pedantic -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-xml.c
gcc -Wall -pedantic -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-web.c
nl-web.c: In function 'getPutPostDeleteUrl':
nl-web.c:338: warning: implicit declaration of function 'alloca'
nl-web.c:338: warning: incompatible implicit declaration of built-in function 'alloca'
nl-web.c: In function 'handleHTTPcgi':
nl-web.c:1250: warning: incompatible implicit declaration of built-in function 'alloca'
gcc -Wall -pedantic -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-matrix.c
gcc -Wall -pedantic -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-debug.c
gcc -Wall -pedantic -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-utf8.c
gcc -Wall -pedantic -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX pcre.c
#gcc newlisp.o nl-symbol.o nl-math.o nl-list.o nl-liststr.o nl-string.o nl-filesys.o nl-sock.o nl-import.o nl-xml.o nl-web.o nl-matrix.o nl-debug.o nl-utf8.o pcre.o -lm -ldl -lrt -lsocket -lnsl -o newlisp
gcc newlisp.o nl-symbol.o nl-math.o nl-list.o nl-liststr.o nl-string.o nl-filesys.o nl-sock.o nl-import.o nl-xml.o nl-web.o nl-matrix.o nl-debug.o nl-utf8.o pcre.o -lm -ldl -lrt -lnsl -o newlisp
ld: 0711-317 ERROR: Undefined symbol: .vasprintf
ld: 0711-345 Use the -bloadmap or -bnoquiet option to obtain more information.
collect2: ld returned 8 exit status
gmake[2]: *** [default] Error 1
gmake[2]: Leaving directory `/tmp/newlisp-10.0.1'
gmake[1]: *** [aix] Error 2
gmake[1]: Leaving directory `/tmp/newlisp-10.0.1'
gmake: *** [default] Error 2
#############################################################################################
I have then tried it using AIX compiler xlc_r by changing 'makefile_aix_utf8' as follows:
#############################################################################################
# makefile for newLISP v. 10.x.x on AIX woth UTF-8 support
#
#
OBJS = newlisp.o nl-symbol.o nl-math.o nl-list.o nl-liststr.o nl-string.o nl-filesys.o
nl-sock.o nl-import.o nl-xml.o nl-web.o nl-matrix.o nl-debug.o nl-utf8.o pcre.o
# use following for UTF-8 support and add nl-utf8.o to the OBJS line
CFLAGS = -c -g -O2 -DSUPPORT_UTF8 -DAIX
CC = xlc_r
default: $(OBJS)
$(CC) $(OBJS) -lm -ldl -lrt -lnsl -o newlisp
.c.o:
$(CC) $(CFLAGS) $<
$(OBJS): primes.h protos.h makefile_aix_utf8
#############################################################################################
I get now
#############################################################################################
...............
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-import.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-xml.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-web.c
"nl-web.c", line 338.10: 1506-068 (W) Operation between types "char*" and "int" is not allowed.
"nl-web.c", line 339.6: 1506-068 (W) Operation between types "char*" and "int" is not allowed.
"nl-web.c", line 340.7: 1506-068 (W) Operation between types "char*" and "int" is not allowed.
"nl-web.c", line 341.6: 1506-068 (W) Operation between types "char*" and "int" is not allowed.
"nl-web.c", line 1250.9: 1506-068 (W) Operation between types "char*" and "int" is not allowed.
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-matrix.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-debug.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-utf8.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX pcre.c
#xlc_r newlisp.o nl-symbol.o nl-math.o nl-list.o nl-liststr.o nl-string.o nl-filesys.o nl-sock.o nl-import.o nl-xml.o nl-web.o nl-matrix.o nl-debug.o nl-utf8.o pcre.o -lm -ldl -lrt -lsocket -lnsl -o newlisp
xlc_r newlisp.o nl-symbol.o nl-math.o nl-list.o nl-liststr.o nl-string.o nl-filesys.o nl-sock.o nl-import.o nl-xml.o nl-web.o nl-matrix.o nl-debug.o nl-utf8.o pcre.o -lm -ldl -lrt -lnsl -o newlisp
ld: 0711-317 ERROR: Undefined symbol: .alloca
ld: 0711-317 ERROR: Undefined symbol: .vasprintf
ld: 0711-345 Use the -bloadmap or -bnoquiet option to obtain more information.
gmake[2]: *** [default] Error 8
gmake[2]: Leaving directory `/tmp/newlisp-10.0.1'
gmake[1]: *** [aix] Error 2
gmake[1]: Leaving directory `/tmp/newlisp-10.0.1'
gmake: *** [default] Error 2
#############################################################################################
Lets take it step by step:
(1) you should definitely stay with gcc not xlc_r
(2) in newlisp.h where you define AIX change to:
#ifdef AIX
#define OSTYPE "AIX"
#define SOLARIS /* <-- this line added */
#endif
put this around line 39 after the #ifdef SOLARIS definition. This makes AIX a sub flavor of SOLARIS (System V, BSD derived)
(3) in newlisp.h after all the includes around line 115 before the #ifndef SOLARIS put:
#ifdef AIX
#define vasprintf my_vasprintf
#endif
After the compiling and linking well do a:
./newlisp qa-dot
inside the newlisp-10.0.1 directory. This is a basic test suite trying out most built-in functions. Most likely we will have problems with system level functions located in nl-filesys.c. If that goes well do a:
make test
in the same location. This runs a more comprehensive test suite.
ps: thanks to nodep for helping out here too.
... also please compile util/types.c and post the output:
> cd newlisp-10.0.1/util
> gcc types.c -o types
> ./types
It will probably give something like this:
type bytes
---------------
char 1
char * 4
void * 4
short int 2
int 4
long 4
long int 4
long long 8
size_t 4
float 4
double 8
long double 16
wchar_t 4
format input output
------------------------------------------------
%d 0xffffffff -1
%u 0xffffffff 4294967295
%d 0x7fffffff 2147483647
%u 0x7fffffff 2147483647
%d 0x80000000 -2147483648
%u 0x80000000 2147483648
%d 0x7fffffffffffffffLL -1
%u 0x7fffffffffffffffLL 4294967295
%x 0xffffffffffffffffLL ffffffff
%X 0x7fffffffffffffffLL ffffffff
%X 0x8000000000000000LL 0
%llX 0xFFFFFFFFFFFFFFFFLL ffffffff
%llX 0x7FFFFFFFFFFFFFFFLL ffffffff
%llX 0x8000000000000000LL 0
%lld 0xffffffffffffffffLL -1
%llu 0xffffffffffffffffLL 18446744073709551615
%lld 0x7fffffffffffffffLL 9223372036854775807
%llu 0x7fffffffffffffffLL 9223372036854775807
%lld 0x8000000000000000LL -9223372036854775808
%llu 0x8000000000000000LL 9223372036854775808
CPU is little endian
The constant __LITTLE_ENDIAN__ is defined by the compiler
The only difference is probably, that your machine may be big endian.
Here is the relevant part pf my "newlisp.h" file:
####################################################################################
+1 /* newlisp.h - header file for newLISP
+2
+3 Copyright (C) 2009 Lutz Mueller
+4
+5 This program is free software: you can redistribute it and/or modify
+6 it under the terms of the GNU General Public License as published by
+7 the Free Software Foundation, either version 3 of the License, or
+8 (at your option) any later version.
+9
+10 This program is distributed in the hope that it will be useful,
+11 but WITHOUT ANY WARRANTY; without even the implied warranty of
+12 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+13 GNU General Public License for more details.
+14
+15 You should have received a copy of the GNU General Public License
+16 along with this program. If not, see <http>.
+17
+18 */
+19
+20 #ifdef LINUX
+21 #define OSTYPE "Linux"
+22 #endif
+23
+24 #ifdef _BSD
+25 #define OSTYPE "BSD"
+26 #endif
+27
+28 #ifdef MAC_OSX
+29 #define OSTYPE "OSX"
+30 #endif
+31
+32 #ifdef SOLARIS
+33 #ifdef TRU64
+34 #define OSTYPE "Tru64Unix"
+35 #else
+36 #define OSTYPE "Solaris"
+37 #endif
+38 #endif
+39
+40 #ifdef AIX
+41 #define OSTYPE "AIX"
+42 #define SOLARIS
+43 #endif
+44
+45 #ifdef WIN_32
+46 #define OSTYPE "Win32"
+47 #endif
+48
+49 #ifdef OS2
+50 #define OSTYPE "OS/2"
+51 #endif
..........
+107 #include <unistd>
+108 #include <sys>
+109 #include <math>
+110 #include <float>
+111 #include <string>
+112 #include <ctype>
+113 #include <fcntl>
+114 #include <stdarg>
+115 #include <time>
+116 #include <sys>
+117 #include <sys>
+118 #include <sys>
+119
+120 #ifdef AIX
+121 #define vasprintf my_vasprintf
+122 #endif
+123
+124 #ifndef SOLARIS
+125 #ifndef _BSD
+126 #ifndef MAC_OSX
+127 #include <malloc>
+128 #endif
+129 #endif
+130 #else
+131 #include <alloca>
+132 #endif
+133
#######################################################################################
I get now the error message:
#gmake
./build
Discovered AIX:
gmake[1]: Entering directory `/tmp/newlisp-10.0.1'
gmake -f makefile_aix_utf8
gmake[2]: Entering directory `/tmp/newlisp-10.0.1'
gcc -Wall -pedantic -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX newlisp.c
newlisp.c:78: error: redefinition of 'opsys'
newlisp.c:73: error: previous definition of 'opsys' was here
gmake[2]: *** [newlisp.o] Error 1
gmake[2]: Leaving directory `/tmp/newlisp-10.0.1'
gmake[1]: *** [aix] Error 2
gmake[1]: Leaving directory `/tmp/newlisp-10.0.1'
gmake: *** [default] Error 2
probably opsys is defined twice in newlisp.c. As of now AIX is a sub-flavor of SOLARIS, similar to TRU64. Change newlisp.c around line 68
#ifdef SOLARIS
int opsys 4
#endif
#ifdef TRU64
int opsys += 5;
#endif
#ifdef AIX
int opsys += 6;
#endif
As both, TRU64 and AIX are sub-flavors of SOLARIS, they will end up with numbers 9 and 10. Then test using the qa-files from 10.0.2, released later today, which elminate all references to opsys numbers from 'sys-info'. It would also help if you can give the output of types.c (see my last post).
After modifying "newlisp.c" as follows
#ifdef SOLARIS
#ifdef TRU64
int opsys = 9;
#endif
#ifdef AIX
int opsys = 10;
#else
int opsys = 4;
#endif
#endif
#################################
I have managed to compile newlisp:-)
Here is the output of gmake:
$ gmake
./build
Discovered AIX:
gmake[1]: Entering directory `/tmp/newlisp-10.0.1'
gmake -f makefile_aix_utf8
gmake[2]: Entering directory `/tmp/newlisp-10.0.1'
gcc -Wall -pedantic -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX newlisp.c
gcc -Wall -pedantic -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-symbol.c
gcc -Wall -pedantic -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-math.c
gcc -Wall -pedantic -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-list.c
gcc -Wall -pedantic -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-liststr.c
gcc -Wall -pedantic -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-string.c
nl-string.c: In function 'p_search':
nl-string.c:135: warning: dereferencing type-punned pointer will break strict-aliasing rules
nl-string.c: In function 'p_pack':
nl-string.c:1619: warning: dereferencing type-punned pointer will break strict-aliasing rules
gcc -Wall -pedantic -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-filesys.c
nl-filesys.c:28:1: warning: "FIONREAD" redefined
In file included from /usr/include/sys/stropts.h:33,
from /usr/include/stropts.h:30,
from nl-filesys.c:26:
/usr/include/sys/ioctl.h:356:1: warning: this is the location of the previous definition
gcc -Wall -pedantic -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-sock.c
nl-sock.c:64:1: warning: "FIONREAD" redefined
In file included from nl-sock.c:47:
/usr/include/sys/ioctl.h:356:1: warning: this is the location of the previous definition
nl-sock.c: In function 'ping':
nl-sock.c:2053: warning: passing argument 6 of 'nrecvfrom' from incompatible pointer type
gcc -Wall -pedantic -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-import.c
gcc -Wall -pedantic -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-xml.c
gcc -Wall -pedantic -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-web.c
gcc -Wall -pedantic -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-matrix.c
gcc -Wall -pedantic -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-debug.c
gcc -Wall -pedantic -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-utf8.c
gcc -Wall -pedantic -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX pcre.c
#gcc newlisp.o nl-symbol.o nl-math.o nl-list.o nl-liststr.o nl-string.o nl-filesys.o nl-sock.o nl-import.o nl-xml.o nl-web.o nl-matrix.o nl-debug.o nl-utf8.o pcre.o -lm -ldl -lrt -lsocket -lnsl -o newlisp
gcc newlisp.o nl-symbol.o nl-math.o nl-list.o nl-liststr.o nl-string.o nl-filesys.o nl-sock.o nl-import.o nl-xml.o nl-web.o nl-matrix.o nl-debug.o nl-utf8.o pcre.o -lm -ldl -lrt -lnsl -o newlisp
gmake[2]: Leaving directory `/tmp/newlisp-10.0.1'
gmake[1]: Leaving directory `/tmp/newlisp-10.0.1'
#########################################################################################
And here is the output of "types":
$ cd util
$ gcc types.c -o types
$./types
type bytes
---------------
char 1
char * 4
void * 4
short int 2
int 4
long 4
long int 4
long long 8
size_t 4
float 4
double 8
long double 8
wchar_t 2
format input output
------------------------------------------------
%d 0xffffffff -1
%u 0xffffffff 4294967295
%d 0x7fffffff 2147483647
%u 0x7fffffff 2147483647
%d 0x80000000 -2147483648
%u 0x80000000 2147483648
%d 0x7fffffffffffffffLL 2147483647
%u 0x7fffffffffffffffLL 2147483647
%x 0xffffffffffffffffLL ffffffff
%X 0x7fffffffffffffffLL 7fffffff
%X 0x8000000000000000LL 80000000
%llX 0xFFFFFFFFFFFFFFFFLL ffffffff
%llX 0x7FFFFFFFFFFFFFFFLL 7fffffff
%llX 0x8000000000000000LL 80000000
%lld 0xffffffffffffffffLL -1
%llu 0xffffffffffffffffLL 18446744073709551615
%lld 0x7fffffffffffffffLL 9223372036854775807
%llu 0x7fffffffffffffffLL 9223372036854775807
%lld 0x8000000000000000LL -9223372036854775808
%llu 0x8000000000000000LL 9223372036854775808
CPU is big endian
The constant __BIG_ENDIAN__ is defined by the compiler
#######################################################
What should the following command do on AIX?
./newlisp qa-dot
Testing built-in functions ...
---------- testing UDP Win32 and OS/2 -------------
waiting ...
sending ...
receiving ...
msg:("HELLO" "127.0.0.1" 42935)
deleting semaphore:true
------------------------------------------
-> share
#######################################################
It is hanging here!!
The types output looks fine, as we don't care about the 'long double' type.
(1) add -fno-strict-aliasing to the make file as done in makefile_solaris
(2) in nl-filesys.c change:
#ifdef SOLARIS
#include <stropts>
#ifndef TRU64
#define FIONREAD I_NREAD
#endif
#endif
to
#ifdef SOLARIS
#include <stropts>
#if !defined (TRU64) || (AIX)
#define FIONREAD I_NREAD
#endif
#endif
For the other warnings I would have to see the header files on your machine in /usr/include/. Some of the functions may have different call-type patterns. If we are lucky it can just be re-mended with a type-cast. The fact that these are only warnings means that we are probably Ok.
In qa-dot replace the definition of test-share with:
(define (test-share) true)
And you make have to do the same with (test-semaphore). One or both of newLISP 'share' and 'semaphore' functions do not work.
This effectively skips the test. We would have to consult the manpages for mmap() and munmap() to find out how memory sharing is done on your system and the same is true for the semaphore() call.
I just see on the snippet, where its hanging on 'share' qa-dot seems to think its in Win32 or OS2. Make sure you are using qa-dot from 10.0.2. After doing the changes in my last post and using 10.0.2-qa-dot, perhaps it will pass 'share' and 'semaphore'.
As you suggested, in files "nl-filesys.c" and "nl-sock.c" I replaced
by
Here is the output of "gmake"
./build
Discovered AIX:
gmake[1]: Entering directory `/tmp/newlisp-10.0.1'
gmake -f makefile_aix_utf8
gmake[2]: Entering directory `/tmp/newlisp-10.0.1'
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX newlisp.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-symbol.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-math.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-list.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-liststr.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-string.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-filesys.c
nl-filesys.c:28:1: warning: "FIONREAD" redefined
In file included from /usr/include/sys/stropts.h:33,
from /usr/include/stropts.h:30,
from nl-filesys.c:26:
/usr/include/sys/ioctl.h:356:1: warning: this is the location of the previous definition
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-sock.c
nl-sock.c:64:1: warning: "FIONREAD" redefined
In file included from nl-sock.c:47:
/usr/include/sys/ioctl.h:356:1: warning: this is the location of the previous definition
nl-sock.c: In function 'ping':
nl-sock.c:2053: warning: passing argument 6 of 'nrecvfrom' from incompatible pointer type
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-import.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-xml.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-web.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-matrix.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-debug.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-utf8.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX pcre.c
#gcc newlisp.o nl-symbol.o nl-math.o nl-list.o nl-liststr.o nl-string.o nl-filesys.o nl-sock.o nl-import.o nl-xml.o nl-web.o nl-matrix.o nl-debug.o nl-utf8.o pcre.o -lm -ldl -lrt -lsocket -lnsl -o newlisp
# OBJECT_MODE=64 gcc newlisp.o nl-symbol.o nl-math.o nl-list.o nl-liststr.o nl-string.o nl-filesys.o nl-sock.o nl-import.o nl-xml.o nl-web.o nl-matrix.o nl-debug.o nl-utf8.o pcre.o -lm -ldl -lrt -lnsl -o newlisp
gcc newlisp.o nl-symbol.o nl-math.o nl-list.o nl-liststr.o nl-string.o nl-filesys.o nl-sock.o nl-import.o nl-xml.o nl-web.o nl-matrix.o nl-debug.o nl-utf8.o pcre.o -lm -ldl -lrt -lnsl -o newlisp
gmake[2]: Leaving directory `/tmp/newlisp-10.0.1'
gmake[1]: Leaving directory `/tmp/newlisp-10.0.1'
Using "qa-dot" of the version 10.0.2 I get now the result:
Testing built-in functions ...
Testing contexts as objects and scoping rules ...
total time: 5564
>>>>> TESTING: ./newlisp FINISHED WITH ERRORS:
>>>> net-accept failed nil
>>>> net-close failed nil
>>>> net-local failed nil
>>>> net-lookup failed nil
>>>> net-peek failed nil
>>>> net-peer failed nil
>>>> net-receive failed nil
>>>> net-sessions failed nil
>>>> share failed ERR: not enough memory in function share
called from user defined function QA:unix-test-share
called from user defined function QA:qa
UTF-8 upper-case: failed
UTF-8 lower-case: failed
Testing contexts as objects and scoping rules ...
total time: 7462
>>>>> TESTING: ./newlisp FINISHED WITH ERRORS:
>>>> net-accept failed nil
>>>> net-close failed nil
>>>> net-local failed nil
>>>> net-lookup failed nil
>>>> net-peek failed nil
>>>> net-peer failed nil
>>>> net-receive failed nil
>>>> net-sessions failed nil
>>>> share failed ERR: not enough memory in function share
called from user defined function QA:unix-test-share
called from user defined function QA:qa
UTF-8 upper-case: failed
UTF-8 lower-case: failed
If it helps you to fix the warnings, I can upload the archive file containing "/usr/include" of my machine.
Compiling newlisp using xlc_r also works smoothly!
./build
Discovered AIX:
gmake[1]: Entering directory `/tmp/newlisp-10.0.1'
gmake -f makefile_aix_utf8
gmake[2]: Entering directory `/tmp/newlisp-10.0.1'
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX newlisp.c
724 1500-010: (W) WARNING in main: Infinite loop. Program may not stop.
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-symbol.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-math.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-list.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-liststr.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-string.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-filesys.c
"nl-filesys.c", line 28.9: 1506-236 (W) Macro name FIONREAD has been redefined.
"nl-filesys.c", line 28.9: 1506-358 (I) "FIONREAD" is defined on line 356 of /usr/include/sys/ioctl.h.
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-sock.c
"nl-sock.c", line 64.9: 1506-236 (W) Macro name FIONREAD has been redefined.
"nl-sock.c", line 64.9: 1506-358 (I) "FIONREAD" is defined on line 356 of /usr/include/sys/ioctl.h.
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-import.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-xml.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-web.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-matrix.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-debug.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-utf8.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX pcre.c
#xlc_r newlisp.o nl-symbol.o nl-math.o nl-list.o nl-liststr.o nl-string.o nl-filesys.o nl-sock.o nl-import.o nl-xml.o nl-web.o nl-matrix.o nl-debug.o nl-utf8.o pcre.o -lm -ldl -lrt -lsocket -lnsl -o newlisp
# OBJECT_MODE=64 xlc_r newlisp.o nl-symbol.o nl-math.o nl-list.o nl-liststr.o nl-string.o nl-filesys.o nl-sock.o nl-import.o nl-xml.o nl-web.o nl-matrix.o nl-debug.o nl-utf8.o pcre.o -lm -ldl -lrt -lnsl -o newlisp
xlc_r newlisp.o nl-symbol.o nl-math.o nl-list.o nl-liststr.o nl-string.o nl-filesys.o nl-sock.o nl-import.o nl-xml.o nl-web.o nl-matrix.o nl-debug.o nl-utf8.o pcre.o -lm -ldl -lrt -lnsl -o newlisp
gmake[2]: Leaving directory `/tmp/newlisp-10.0.1'
gmake[1]: Leaving directory `/tmp/newlisp-10.0.1'
"newlisp qa-dot" returns the expected result as before
Testing built-in functions ...
Testing contexts as objects and scoping rules ...
total time: 8559
>>>>> TESTING: ./newlisp FINISHED WITH ERRORS:
>>>> net-accept failed nil
>>>> net-close failed nil
>>>> net-local failed nil
>>>> net-lookup failed nil
>>>> net-peek failed nil
>>>> net-peer failed nil
>>>> net-receive failed nil
>>>> net-sessions failed nil
>>>> share failed ERR: not enough memory in function share
called from user defined function QA:unix-test-share
called from user defined function QA:qa
UTF-8 upper-case: failed
UTF-8 lower-case: failed
Testing contexts as objects and scoping rules ...
total time: 11004
>>>>> TESTING: ./newlisp FINISHED WITH ERRORS:
>>>> net-accept failed nil
>>>> net-close failed nil
>>>> net-local failed nil
>>>> net-lookup failed nil
>>>> net-peek failed nil
>>>> net-peer failed nil
>>>> net-receive failed nil
>>>> net-sessions failed nil
>>>> share failed ERR: not enough memory in function share
called from user defined function QA:unix-test-share
called from user defined function QA:qa
UTF-8 upper-case: failed
UTF-8 lower-case: failed
I have also tried to compile a 64-BIT version of newlisp using the following makefile:
nl-sock.o nl-import.o nl-xml.o nl-web.o nl-matrix.o nl-debug.o nl-utf8.o pcre.o
# use following for UTF-8 support and add nl-utf8.o to the OBJS line
CFLAGS = -q64 -c -g -O2 -DSUPPORT_UTF8 -DAIX
CC = xlc_r
default: $(OBJS)
OBJECT_MODE=64 $(CC) $(OBJS) -lm -ldl -lrt -lnsl -o newlisp
.c.o:
$(CC) $(CFLAGS) $<
$(OBJS): primes.h protos.h makefile_aix_utf8
Here is the output of gmake:
./build
Discovered AIX:
gmake[1]: Entering directory `/tmp/newlisp-10.0.1'
gmake -f makefile_aix_utf8
gmake[2]: Entering directory `/tmp/newlisp-10.0.1'
xlc_r -q64 -c -g -O2 -DSUPPORT_UTF8 -DAIX newlisp.c
724 1500-010: (W) WARNING in main: Infinite loop. Program may not stop.
xlc_r -q64 -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-symbol.c
xlc_r -q64 -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-math.c
xlc_r -q64 -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-list.c
xlc_r -q64 -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-liststr.c
xlc_r -q64 -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-string.c
xlc_r -q64 -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-filesys.c
"nl-filesys.c", line 28.9: 1506-236 (W) Macro name FIONREAD has been redefined.
"nl-filesys.c", line 28.9: 1506-358 (I) "FIONREAD" is defined on line 356 of /usr/include/sys/ioctl.h.
xlc_r -q64 -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-sock.c
"nl-sock.c", line 64.9: 1506-236 (W) Macro name FIONREAD has been redefined.
"nl-sock.c", line 64.9: 1506-358 (I) "FIONREAD" is defined on line 356 of /usr/include/sys/ioctl.h.
xlc_r -q64 -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-import.c
xlc_r -q64 -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-xml.c
xlc_r -q64 -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-web.c
xlc_r -q64 -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-matrix.c
xlc_r -q64 -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-debug.c
xlc_r -q64 -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-utf8.c
xlc_r -q64 -c -g -O2 -DSUPPORT_UTF8 -DAIX pcre.c
#xlc_r newlisp.o nl-symbol.o nl-math.o nl-list.o nl-liststr.o nl-string.o nl-filesys.o nl-sock.o nl-import.o nl-xml.o nl-web.o nl-matrix.o nl-debug.o nl-utf8.o pcre.o -lm -ldl -lrt -lsocket -lnsl -o newlisp
OBJECT_MODE=64 xlc_r newlisp.o nl-symbol.o nl-math.o nl-list.o nl-liststr.o nl-string.o nl-filesys.o nl-sock.o nl-import.o nl-xml.o nl-web.o nl-matrix.o nl-debug.o nl-utf8.o pcre.o -lm -ldl -lrt -lnsl -o newlisp
#xlc_r newlisp.o nl-symbol.o nl-math.o nl-list.o nl-liststr.o nl-string.o nl-filesys.o nl-sock.o nl-import.o nl-xml.o nl-web.o nl-matrix.o nl-debug.o nl-utf8.o pcre.o -lm -ldl -lrt -lnsl -o newlisp
gmake[2]: Leaving directory `/tmp/newlisp-10.0.1'
gmake[1]: Leaving directory `/tmp/newlisp-10.0.1'
But this time the test results in a segmentation fault:
Testing built-in functions ...
Segmentation fault (core dumped)
I have done the following naive check, whether the newlisp works at all:
newLISP v.10.0.1 on AIX IPv4 UTF-8, execute 'newlisp -h' for more info.
> (define fac (lambda (n) (if (< n 1) 1 (* n (fac (- n 1))))))
(lambda (n)
(if (< n 1)
1
(* n (fac (- n 1)))))
> (fac 10)
3628800
>
It does!
How can I compile a 64-BIT version of newlisp using gcc?
Looks like FIONREAD is already defined in ioctl.h and perhaps we don't have to include stropts.h either.
In the original nl-filesys.c:
#include <stropts>
#ifndef TRU64
#define FIONREAD I_NREAD
#endif
#endif
change to
#ifdef SOLARIS
#include <stropts>
#ifndef TRU64
#define FIONREAD I_NREAD
#endif
#endif
#endif
and in nl-sock.c:
#include <stropts>
#include <sys>
#include <netinet>
#define gethostbyname2(A, B) gethostbyname(A)
#ifndef TRU64
#define FIONREAD I_NREAD
#endif
#endif
change to:
#ifndef AIX
#include <stropts>
#endif
#include <sys>
#include <netinet>
#define gethostbyname2(A, B) gethostbyname(A)
#ifndef TRU64
#ifndef AIX
#define FIONREAD I_NREAD
#endif
#endif
#endif
This is getting all very messy. Once we have it working I will detangle all the SOLARIS dependencies and make TRU64 and AIX independent flavors. But for now lets get it working first.
The above would only fix the peek and net-peek functiona, which are the only once using that special FIONREAD or I_NREAD flag.
The fact that all of the net-xxx functions fail, tells me that something more fundamental is failing, because most of the network code is completely portable, even between Unix and Win32.
As if some special initialization for the socket-, network- function stack is necessary on AIX?
Perhaps the man pages for the socket() call gives us som info about this. Is this AIX installation very old? From what year is it?
In the makefile put -m32 at the beginning of the CFLAGS line and the beginning of the $(CC) line for gcc. Although you make have a 64-bit machine I want to compile a 32-bit API first. If this is not possible on your machine (It should) then you have have to add the flag -DNEWLISP64 in the CFLAGS line, and you can add -m64 to the gcc CFLAGS and $(CC) lines. -DNEWLISP64 should go in for any compiler when compiling for LP64 model.
ps: note that in all #include's the forum software has stripped the dot-h after the filenames.
Although there are still some open issues on AIX platform, I would like
to summarize the necessary steps for compiling newLISP on AIX.
1) Depending on your C compiler you should use one of the following make files:
GNU C Compiler (gcc)
--------------------
#============================ makefile_aix_utf8 (gcc) ==================================
#
# makefile for newLISP v. 10.x.x on AIX with UTF-8 support on PowerPC CPU
#
OBJS = newlisp.o nl-symbol.o nl-math.o nl-list.o nl-liststr.o nl-string.o nl-filesys.o
nl-sock.o nl-import.o nl-xml.o nl-web.o nl-matrix.o nl-debug.o nl-utf8.o pcre.o
# use following for UTF-8 support and add nl-utf8.o to the OBJS line
CFLAGS = -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX
CC = gcc
default: $(OBJS)
$(CC) $(OBJS) -lm -ldl -lrt -lnsl -o newlisp
.c.o:
$(CC) $(CFLAGS) $<
$(OBJS): primes.h protos.h makefile_aix_utf8
#=====================================================================================
IBM C Compiler (xlc_r)
----------------------
#========================= makefile_aix_utf8 (xlc_r) =================================
#
# makefile for newLISP v. 10.x.x on AIX with UTF-8 support on PowerPC CPU
#
#
OBJS = newlisp.o nl-symbol.o nl-math.o nl-list.o nl-liststr.o nl-string.o nl-filesys.o
nl-sock.o nl-import.o nl-xml.o nl-web.o nl-matrix.o nl-debug.o nl-utf8.o pcre.o
# use following for UTF-8 support and add nl-utf8.o to the OBJS line
CFLAGS = -c -g -O2 -DSUPPORT_UTF8 -DAIX
CC = xlc_r
default: $(OBJS)
$(CC) $(OBJS) -lm -ldl -lrt -lnsl -o newlisp
.c.o:
$(CC) $(CFLAGS) $<
$(OBJS): primes.h protos.h makefile_aix_utf8
#=====================================================================================
and for 64bit version of newLISP
#
# makefile for newLISP v. 10.x.x on AIX with UTF-8 support on PowerPC CPU
#
#
OBJS = newlisp.o nl-symbol.o nl-math.o nl-list.o nl-liststr.o nl-string.o nl-filesys.o
nl-sock.o nl-import.o nl-xml.o nl-web.o nl-matrix.o nl-debug.o nl-utf8.o pcre.o
# use following for UTF-8 support and add nl-utf8.o to the OBJS line
CFLAGS = -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX
CC = xlc_r
default: $(OBJS)
OBJECT_MODE=64 $(CC) $(OBJS) -lm -ldl -lrt -lnsl -o newlisp
.c.o:
$(CC) $(CFLAGS) $<
$(OBJS): primes.h protos.h makefile_aix_utf8
#=====================================================================================
2) In "newlisp.h" insert the following lines after the line 122.
befor insert:
+122 #include <sys/types.h>
+123
+124 #ifndef SOLARIS
+125 #ifndef _BSD
+126 #ifndef MAC_OSX
+127 #include <malloc.h>
+128 #endif
+129 #endif
+130 #else
+131 #include <alloca.h>
+132 #endif
after insert:
+122 #include <sys/types.h>
+123
+124 #ifdef AIX
+125 #define vasprintf my_vasprintf
+126 #endif
+127
+128 #ifndef SOLARIS
+129 #ifndef _BSD
+130 #ifndef MAC_OSX
+131 #include <malloc.h>
+132 #endif
+133 #endif
+134 #else
+135 #include <alloca.h>
+136 #endif
3) In "newlisp.c" replace
+70 int opsys 4
+71 #endif
+72
+73 #ifdef TRU64
+74 int opsys += 5;
+75 #endif
+76
+77 #ifdef AIX
+78 int opsys += 6;
+79 #endif
with
+70 #ifdef TRU64
+71 int opsys = 9;
+72 #endif
+73 #ifdef AIX
+74 int opsys = 10;
+75 #else
+76 int opsys = 4;
+77 #endif
+78 #endif
4) In "nl-filesystem.c" replace
+26 #include <stropts.h>
+27 #ifndef TRU64
+28 #define FIONREAD I_NREAD
+29 #endif
+30 #endif
with
+26 #include <stropts.h>
+27 #ifndef TRU64
+28 #ifndef AIX
+29 #define FIONREAD I_NREAD
+30 #endif
+31 #endif
+32 #endif
5) In "nl-sock.c" replace
+61 #include <stropts.h>
+62 #include <sys/conf.h>
+63 #include <netinet/in_systm.h>
+64 #define gethostbyname2(A, B) gethostbyname(A)
+65 #ifndef TRU64
+66 #define FIONREAD I_NREAD
+67 #endif
+68 #endif
with
+61 #include <stropts.h>
+62 #include <sys/conf.h>
+63 #include <netinet/in_systm.h>
+64 #define gethostbyname2(A, B) gethostbyname(A)
+65 #ifndef TRU64
+66 #ifndef AIX
+67 #define FIONREAD I_NREAD
+68 #endif
+69 #endif
+70 #endif
6) In "Makefile" change all occurrences of make to gmake and insert an entry for AIX make file
after e.g. opensolaris entry:
+143 gmake -f makefile_opensolaris
+144
+145 aix:
+146 gmake -f makefile_aix_utf8
+147
+148 mingw:
+149 gmake -f makefile_mingw
7) in "build" change all occurrences of make to gmake and insert an entry for AIX
+43 echo Discovered SunOS, making for Sparc CPU:
+44 gmake solaris
+45 exit
+46 ;;
+47 AIX)
+48 echo Discovered AIX:
+49 gmake aix
+50 exit
+51 ;;
+52 OSF1)
+53 echo Discovered Tru64 Unix, please read doc/TRU64BUILD
+54 gmake -f gmakefile_tru64
+55 exit
+56 ;;
And you see here the output of compilation runs on my machine
(AIX 5.3 CPU type:PowerPC_POWER5 1.9 GHz):
IBM C Compiler (xlc_r 64bit)
----------------------------
# gmake
./build
Discovered AIX:
gmake[1]: Entering directory `/tmp/newlisp-10.0.2'
gmake -f makefile_aix_utf8
gmake[2]: Entering directory `/tmp/newlisp-10.0.2'
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX newlisp.c
728 1500-010: (W) WARNING in main: Infinite loop. Program may not stop.
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-symbol.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-math.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-list.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-liststr.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-string.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-filesys.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-sock.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-import.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-xml.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-web.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-matrix.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-debug.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-utf8.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX pcre.c
OBJECT_MODE=64 xlc_r newlisp.o nl-symbol.o nl-math.o nl-list.o nl-liststr.o nl-string.o nl-filesys.o nl-sock.o nl-import.o nl-xml.o nl-web.o nl-matrix.o nl-debug.o nl-utf8.o pcre.o -lm -ldl -lrt -lnsl -o newlisp
gmake[2]: Leaving directory `/tmp/newlisp-10.0.2'
gmake[1]: Leaving directory `/tmp/newlisp-10.0.2'
The environment variable OBJECT_MODE=64 is necessary for 64bit version, otherwise you get the following linking error:
XCOFF64 object files are not allowed in 32-bit mode.
Testing built-in functions ...
Testing contexts as objects and scoping rules ...
total time: 8458
>>>>> TESTING: ./newlisp FINISHED WITH ERRORS:
>>>> net-lookup failed nil
UTF-8 upper-case: failed
UTF-8 lower-case: failed
GNU C Compiler (gcc)
--------------------
# gmake
./build
Discovered AIX:
gmake[1]: Entering directory `/tmp/newlisp-10.0.2'
gmake -f makefile_aix_utf8
gmake[2]: Entering directory `/tmp/newlisp-10.0.2'
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX newlisp.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-symbol.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-math.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-list.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-liststr.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-string.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-filesys.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-sock.c
nl-sock.c: In function 'ping':
nl-sock.c:2060: warning: passing argument 6 of 'nrecvfrom' from incompatible pointer type
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-import.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-xml.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-web.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-matrix.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-debug.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-utf8.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX pcre.c
gcc newlisp.o nl-symbol.o nl-math.o nl-list.o nl-liststr.o nl-string.o nl-filesys.o nl-sock.o nl-import.o nl-xml.o nl-web.o nl-matrix.o nl-debug.o nl-utf8.o pcre.o -lm -ldl -lrt -lnsl -o newlisp
gmake[2]: Leaving directory `/tmp/newlisp-10.0.2'
gmake[1]: Leaving directory `/tmp/newlisp-10.0.2'
# ./newlisp qa-dot
Testing built-in functions ...
Testing contexts as objects and scoping rules ...
total time: 5477
>>>>> TESTING: ./newlisp FINISHED WITH ERRORS:
>>>> net-lookup failed nil
>>>> share failed ERR: not enough memory in function share
called from user defined function QA:unix-test-share
called from user defined function QA:qa
UTF-8 upper-case: failed
UTF-8 lower-case: failed
Testing contexts as objects and scoping rules ...
total time: 7364
>>>>> TESTING: ./newlisp FINISHED WITH ERRORS:
>>>> net-lookup failed nil
>>>> share failed ERR: not enough memory in function share
called from user defined function QA:unix-test-share
called from user defined function QA:qa
UTF-8 upper-case: failed
UTF-8 lower-case: failed
As you see there are still some failing tests and a compilation warning.
By the way just for your information gcc doesn't accept -m64 on AIX!
./build
Discovered AIX:
gmake[1]: Entering directory `/tmp/newlisp-10.0.2'
gmake -f makefile_aix_utf8
gmake[2]: Entering directory `/tmp/newlisp-10.0.2'
gcc -m64 -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX newlisp.c
cc1: error: invalid option '64'
gmake[2]: *** [newlisp.o] Error 1
gmake[2]: Leaving directory `/tmp/newlisp-10.0.2'
gmake[1]: *** [aix] Error 2
gmake[1]: Leaving directory `/tmp/newlisp-10.0.2'
gmake: *** [default] Error 2
Thanks for summarizing all the steps necessary. I will try to integrate this into the next maintenance or development release and with both makefiles. Separating the AIX and TRU64 flavors from SOLARIS, will make the #ifdef's simpler and less confusing.
The 'net-lookup' function relies on "localhost" beeing defined as 127.0.0.1. May be there is a different setting on your machine?
The UTF-8 upper-case/lower-case failures are probably due to wrong or missing locale settings on your machine. The 'recvfrom' warning using gcc can probably be fixed by casting the last parameter (in or close to line 2059):
... may be for now we include only the makefile for the IBM xlc_r compiler, because of the 'share' error. Unless you are able to find out what's going on there. The fact that the same code works fine using the xlc_r compiler, means perhaps that there is a problem with gcc on AIX? Except for Win32 there are no special platform dependencies in this function. Perhaps it is a flags issue with the mmap() function.
Is the xlc_r compiler included on this machine by default? Does everybody using this system have it?
After implementing your suggested modification
I get now the following outputs
IBM Compiler xlc_r 64bit
------------------------
# gmake
./build
Discovered AIX:
gmake[1]: Entering directory `/tmp/newlisp-10.0.2'
gmake -f makefile_aix_utf8
gmake[2]: Entering directory `/tmp/newlisp-10.0.2'
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX newlisp.c
728 1500-010: (W) WARNING in main: Infinite loop. Program may not stop.
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-symbol.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-math.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-list.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-liststr.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-string.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-filesys.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-sock.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-import.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-xml.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-web.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-matrix.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-debug.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-utf8.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX pcre.c
OBJECT_MODE=64 xlc_r newlisp.o nl-symbol.o nl-math.o nl-list.o nl-liststr.o nl-string.o nl-filesys.o nl-sock.o nl-import.o nl-xml.o nl-web.o nl-matrix.o nl-debug.o nl-utf8.o pcre.o -lm -ldl -lrt -lnsl -o newlisp
gmake[2]: Leaving directory `/tmp/newlisp-10.0.2'
gmake[1]: Leaving directory `/tmp/newlisp-10.0.2'
and
GNU C Compiler (gcc)
--------------------
# gmake
./build
Discovered AIX:
gmake[1]: Entering directory `/tmp/newlisp-10.0.2'
gmake -f makefile_aix_utf8
gmake[2]: Entering directory `/tmp/newlisp-10.0.2'
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX newlisp.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-symbol.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-math.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-list.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-liststr.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-string.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-filesys.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-sock.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-import.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-xml.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-web.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-matrix.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-debug.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-utf8.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX pcre.c
gcc newlisp.o nl-symbol.o nl-math.o nl-list.o nl-liststr.o nl-string.o nl-filesys.o nl-sock.o nl-import.o nl-xml.o nl-web.o nl-matrix.o nl-debug.o nl-utf8.o pcre.o -lm -ldl -lrt -lnsl -o newlisp
gmake[2]: Leaving directory `/tmp/newlisp-10.0.2'
gmake[1]: Leaving directory `/tmp/newlisp-10.0.2'
It looks pretty good:-)
The failure of "net-lookup" test was because of an additional name for "127.0.0.1" in my /etc/hosts file:
One can check it directly:
newLISP v.10.0.2 64-bit on AIX IPv4 UTF-8, execute 'newlisp -h' for more info.
> (net-lookup "localhost")
"127.0.0.1"
> (net-lookup "127.0.0.1")
"loopback"
>
Regarding "UTF-8 upper-case" and "UTF-8 lower-case" tests it seems
that something is wrong with "upper-case" and "lower-case" functions.
Because I'm just a (new)LISP newbie my findings maybe totally wrong!
It seems that these functions act as identity on characters beyond
ASCII range! (I know it doesn't help you. It is just my observation.)
My locale setting:
LANG=en_US
LC_COLLATE="en_US"
LC_CTYPE="en_US"
LC_MONETARY="en_US"
LC_NUMERIC="en_US"
LC_TIME="en_US"
LC_MESSAGES="en_US"
LC_ALL=
And here is the result of my simple experiment using code fragments from qa-dot file:
newLISP v.10.0.2 64-bit on AIX IPv4 UTF-8, execute 'newlisp -h' for more info.
> (= (length (char 937)) 2)
true
> (set-locale "en_US")
("en_US en_US en_US en_US en_US en_US" ".")
> (set 'unicodelist '(913 914 915 916 937 945 946 947 948 969 32 1040 1041 1042 1043 1044 1072 1073 1074 1075 1076 13 10))
(913 914 915 916 937 945 946 947 948 969 32 1040 1041 1042 1043 1044 1072 1073 1074 1075 1076 13 10)
> (set 'utf8str (join (map char unicodelist)))
"ÎÎÎÎÎ©Î±Î²Î³Î´Ï ÐÐÐÐÐабвгдrn"
> (= (map char (explode (upper-case utf8str))) '(913 914 915 916 937 913 914 915 916 937 32 1040 1041 1042 1043 1044 1040 1041 1042 1043 1044 13 10))
nil
> (map char (explode (upper-case utf8str)))
(913 914 915 916 937 945 946 947 948 969 32 1040 1041 1042 1043 1044 1072 1073 1074 1075 1076 13 10)
>
Do you have any idea?
Regarding your question about IBM Compiler: It is unfortunately neither free nor included by default!
Many systems do not contain upper-, lower- case mappings for wide characters which are not from their own locale, even if their compiler offers the necessary towuppper() and towlower() C-functions, that is normal. Your compile will work well for UTF-8 characters otherwise. Only on Mac OS X, UBUNTU Linux and Win32, localized for most of the world, upper- and lower- casing UTF-8 characters beyond the ASCII part works well consistently.
Regarding running 64-bit newLISP, be aware that this doubles memory requirements for many data and operations in newLISP, as all pointers are now 64-bit. On the upside your program runs about 30% faster on average and potentially can use all memory available on your machine. Also, it seems the 64-bit option for gcc on AIX is: -maix64. See here:
http://www.ibm.com/developerworks/aix/library/au-gnu.html
Although IBM xlc_r is not free and installed by default, I assume that most AIX installations have it. At least from googling around on AIX, I have that impression. I will include both makefiles, but move the gcc one in the util directory with a comment about 'share'.
The next version 10.0.3 will have all your changes integrated and with AIX and TRU64 as flavors independent from SOLARIS. AIX will stay with 10 for the 'sys-info' number. If you want, I can make it available for you earlier to test building on AIX.
I have upgraded my gcc installation to version 4.2.4.
Compiling with option -maix64 works but by calling ./newlisp qa-dot
I get segmentation fault. The 32bit version compiled by gcc suffers
from the same problem as before. I would say that you should
integrate just the xlc_r version into the upcoming maintenance
release. The gcc version can be provided as an alternative to xlc_r
version supplied by a remark about issues that are still open.
Thank you very much for pointing out URL. I came to this guide
as I have struggled with compiling Python on AIX:-)
Nearly the same problems (and a few more) as here.
It would give me great pleasure to support you in your effort to
improve newLISP (at least with regard to its portability).
I will do it (I mean testing newLISP 10.0.3 on AIX).
I can also test it on Solaris (SPARC) if you don't have access to it.
Where can I get it?
Regarding gcc, version 4.3.2 seems to be fine on Linux, on Mac OS X I am using now 4.2.1 previously 4.0.1 and they all seem to be fine. It would be good though to get gcc with newLISP going on AIX too.
I left you a private message, how to pickup a 10.0.3 preview. After it tests well on AIX using "make test", I will do either a development release or wait for the maintenance release until May.
Note, that I only changed to "gmake" in the few places where necessary for AIX. As Mac OS X, Win32 and Linux typically refer to it as "make". I hope I covered all necessary places.
For SOLARIS I am using:
"SunOS carbon 5.8 Generic_117350-25 sun4u sparc SUNW,Ultra-2"
Another user on this board made it available for me accessible via the SSH. All releases get fully tested on that machine.
The user pjotl is porting to TRU64 on a Compaq system, which is very non-standard how it implements certain data types. The AIX port so far has been a pretty standard Unix port, thanks to the ANSI compatible IBM xlc_r compiler.
Just for your info
./build
Discovered AIX:
gmake[1]: Entering directory `/tmp/newlisp-10.0.3'
gmake -f makefile_aix_utf8_xlc
gmake[2]: Entering directory `/tmp/newlisp-10.0.3'
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX newlisp.c
729 1500-010: (W) WARNING in main: Infinite loop. Program may not stop.
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-symbol.c
"nl-symbol.c", line 131.7: 1506-068 (W) Operation between types "char*" and "int" is not allowed.
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-math.c
"nl-math.c", line 492.13: 1506-068 (W) Operation between types "char*" and "int" is not allowed.
"nl-math.c", line 493.16: 1506-068 (W) Operation between types "char*" and "int" is not allowed.
"nl-math.c", line 2156.10: 1506-068 (W) Operation between types "struct {...}**" and "int" is not allowed.
"nl-math.c", line 2157.7: 1506-068 (W) Operation between types "int*" and "int" is not allowed.
"nl-math.c", line 2158.7: 1506-068 (W) Operation between types "char*" and "int" is not allowed.
"nl-math.c", line 2358.8: 1506-068 (W) Operation between types "double*" and "int" is not allowed.
"nl-math.c", line 2359.7: 1506-068 (W) Operation between types "double*" and "int" is not allowed.
"nl-math.c", line 2360.15: 1506-068 (W) Operation between types "double*" and "int" is not allowed.
"nl-math.c", line 2365.11: 1506-068 (W) Operation between types "double*" and "int" is not allowed.
"nl-math.c", line 2366.11: 1506-068 (W) Operation between types "double*" and "int" is not allowed.
"nl-math.c", line 2398.7: 1506-068 (W) Operation between types "char*" and "int" is not allowed.
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-list.c
"nl-list.c", line 1377.15: 1506-068 (W) Operation between types "unsigned long*" and "int" is not allowed.
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-liststr.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-string.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-filesys.c
"nl-filesys.c", line 2012.25: 1506-068 (W) Operation between types "char*" and "int" is not allowed.
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-sock.c
"nl-sock.c", line 1947.47: 1506-068 (W) Operation between types "char*" and "int" is not allowed.
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-import.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-xml.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-web.c
"nl-web.c", line 342.10: 1506-068 (W) Operation between types "char*" and "int" is not allowed.
"nl-web.c", line 343.6: 1506-068 (W) Operation between types "char*" and "int" is not allowed.
"nl-web.c", line 344.7: 1506-068 (W) Operation between types "char*" and "int" is not allowed.
"nl-web.c", line 345.6: 1506-068 (W) Operation between types "char*" and "int" is not allowed.
"nl-web.c", line 1274.9: 1506-068 (W) Operation between types "char*" and "int" is not allowed.
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-matrix.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-debug.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-utf8.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX pcre.c
xlc_r newlisp.o nl-symbol.o nl-math.o nl-list.o nl-liststr.o nl-string.o nl-filesys.o nl-sock.o nl-import.o nl-xml.o nl-web.o nl-matrix.o nl-debug.o nl-utf8.o pcre.o -lm -ldl -lrt -lnsl -o newlisp
ld: 0711-317 ERROR: Undefined symbol: .alloca
ld: 0711-345 Use the -bloadmap or -bnoquiet option to obtain more information.
gmake[2]: *** [default] Error 8
gmake[2]: Leaving directory `/tmp/newlisp-10.0.3'
gmake[1]: *** [aix] Error 2
gmake[1]: Leaving directory `/tmp/newlisp-10.0.3'
gmake: *** [default] Error 2
It lost the include of alloca.h. I uploaded a new newlisp-10.0.3.tgz in your place with a changed newlisp.h.
I would change the following line
in both make files to
and do the following renaming in order to be consistent with the entry in Makefile
and then change
in makefile_aix_utf8_xlcLP64 to
Here is the output of compilation runs that looks pretty good:-)
gmake -f makefile_aix_utf8_xlcLP64
gmake[1]: Entering directory `/tmp/newlisp-10.0.3'
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX newlisp.c
729 1500-010: (W) WARNING in main: Infinite loop. Program may not stop.
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-symbol.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-math.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-list.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-liststr.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-string.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-filesys.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-sock.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-import.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-xml.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-web.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-matrix.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-debug.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-utf8.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX pcre.c
OBJECT_MODE=64 xlc_r newlisp.o nl-symbol.o nl-math.o nl-list.o nl-liststr.o nl-string.o nl-filesys.o nl-sock.o nl-import.o nl-xml.o nl-web.o nl-matrix.o nl-debug.o nl-utf8.o pcre.o -lm -ldl -lrt -lnsl -o newlisp
gmake[1]: Leaving directory `/tmp/newlisp-10.0.3'
# ./newlisp qa-dot
Testing built-in functions ...
Testing contexts as objects and scoping rules ...
total time: 5700
>>>>> TESTING: ./newlisp FINISHED WITH ERRORS:
UTF-8 upper-case: failed
UTF-8 lower-case: failed
./build
Discovered AIX:
gmake[1]: Entering directory `/tmp/newlisp-10.0.3'
gmake -f makefile_aix_utf8_xlc
gmake[2]: Entering directory `/tmp/newlisp-10.0.3'
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX newlisp.c
729 1500-010: (W) WARNING in main: Infinite loop. Program may not stop.
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-symbol.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-math.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-list.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-liststr.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-string.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-filesys.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-sock.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-import.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-xml.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-web.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-matrix.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-debug.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-utf8.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX pcre.c
xlc_r newlisp.o nl-symbol.o nl-math.o nl-list.o nl-liststr.o nl-string.o nl-filesys.o nl-sock.o nl-import.o nl-xml.o nl-web.o nl-matrix.o nl-debug.o nl-utf8.o pcre.o -lm -ldl -lrt -lnsl -o newlisp
gmake[2]: Leaving directory `/tmp/newlisp-10.0.3'
gmake[1]: Leaving directory `/tmp/newlisp-10.0.3'
# ./newlisp qa-dot
Testing built-in functions ...
Testing contexts as objects and scoping rules ...
total time: 5522
>>>>> TESTING: ./newlisp FINISHED WITH ERRORS:
>>>> share failed ERR: not enough memory in function share
called from user defined function QA:unix-test-share
called from user defined function QA:qa
UTF-8 upper-case: failed
UTF-8 lower-case: failed
Testing contexts as objects and scoping rules ...
total time: 7435
>>>>> TESTING: ./newlisp FINISHED WITH ERRORS:
>>>> share failed ERR: not enough memory in function share
called from user defined function QA:unix-test-share
called from user defined function QA:qa
UTF-8 upper-case: failed
UTF-8 lower-case: failed
As you see the problem with "share" exists also for xlc_r in 32bit mode.
I would make 64bit the default choice for compilation on AIX for the time being!
Do you have any idea where we should begin to fix this problem?
How critical is it?
I changed to this:
gmake -f makefile_aixLP64_utf8_xlc
in Makefile, as all makefile_xxx put the memory model-id: LP64 after the platform name. I also changed from Sparc to PowerPC in the aix makefiles.
Memory mapping is used in 'share' and in the Cilk multiprocessing API consisting of the functions 'spawn, 'sync' and 'abort'. The Cilk API is an extreme convenient standard for doing multiprocessing. I would love this to work.
To fix this one would have to study the docs/man pages for mmap() and munmap(). The mmap() call takes a load of different flags (modes) which may be platform sensitive and memory-model sensitive. I am pretty confident, that with a little bit of reading and experimentation this can be fixed for xlc_r 32-bit.
Another important parameter in mmap() is the pagesize. In newlisp.c the pagesize is aquired in line 585 with a getpagesize() call.
Here is a little test program to check memory mapping (used for sharing memory) on your machine:
#include <stdio>
#include <sys>
int main(int argc, char ** argv)
{
void * address;
int pagesize;
pagesize = getpagesize();
printf("pagesize: %dn", pagesize);
address = mmap( 0, pagesize, PROT_READ | PROT_WRITE, MAP_SHARED | MAP_ANON, -1, 0);
printf("address:0x%Xn", address);
}
~> gcc -m32 test-mmap.c -o test-mmap
~> ./test-mmap
pagesize: 4096
address:0x8000
~>
You could try different flags and the xlc_r compiler with 32-bit memory model. I have this test program included in the distribution as util/test-mmap.c
Once mmap() works the Cilk API in newLISP is tested using qa-cilk, it launches 41 processes and gives memory mapping a good workout.
Another newlisp-10.0.3.tgz is in the usual place. The only changes I made are in the makefiles and the new test-mmap.c in util/. Thanks for the help! Looks like we are getting there :)
Thank you for implementing the modifications!
I have tested the new version and it works fine. I have also tried the 64bit version of gcc.
It also works fine. Last time I have forgotten to add the option -DNEWLISP64 to CFLAGS and
this was the reason why I got the segmentation fault! That means that gcc is on a par with xlc!
For the sake of completeness I append the make file that I used for gcc (64bit):
# makefile for newLISP v. 10.x.x on AIX with UTF-8 support on PowerPC CPU
#
OBJS = newlisp.o nl-symbol.o nl-math.o nl-list.o nl-liststr.o nl-string.o nl-filesys.o
nl-sock.o nl-import.o nl-xml.o nl-web.o nl-matrix.o nl-debug.o nl-utf8.o pcre.o
# use following for UTF-8 support and add nl-utf8.o to the OBJS line
CFLAGS = -maix64 -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX
CC = gcc
default: $(OBJS)
$(CC) $(OBJS) -maix64 -lm -ldl -lrt -lnsl -o newlisp
.c.o:
$(CC) $(CFLAGS) $<
$(OBJS): primes.h protos.h makefile_aixLP64_utf8_gcc
#
#==============================================================================
And here is the 32bit version for gcc (to facilitate Copy&Paste if necessary):
# makefile for newLISP v. 10.x.x on AIX with UTF-8 support on PowerPC CPU
#
OBJS = newlisp.o nl-symbol.o nl-math.o nl-list.o nl-liststr.o nl-string.o nl-filesys.o
nl-sock.o nl-import.o nl-xml.o nl-web.o nl-matrix.o nl-debug.o nl-utf8.o pcre.o
# use following for UTF-8 support and add nl-utf8.o to the OBJS line
CFLAGS = -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX
CC = gcc
default: $(OBJS)
$(CC) $(OBJS) -lm -ldl -lrt -lnsl -o newlisp
.c.o:
$(CC) $(CFLAGS) $<
$(OBJS): primes.h protos.h makefile_aix_utf8_gcc
#
#==============================================================================
I would suggest that we modify the Makefile such that both compiler types can be used:
aix64:
gmake -f makefile_aixLP64_utf8_xlc
aix:
gmake -f makefile_aix_utf8_xlc
aix64_gcc:
gmake -f makefile_aixLP64_utf8_gcc
aix_gcc:
gmake -f makefile_aix_utf8_gcc
...
or something like this.
Here is the output of test-map for both gcc and xlc (64bit/32bit):
# ./test-mmap
pagesize: 4096
address:0x0
# xlc -q64 test-mmap.c -o test-mmap
# ./test-mmap
pagesize: 4096
address:0x0
# gcc test-mmap.c -o test-mmap
# ./test-mmap
pagesize: 4096
address:0x30000000
# xlc test-mmap.c -o test-mmap
# ./test-mmap
pagesize: 4096
address:0x30000000
I agree with you on Cilk API. I find the language features regarding concurrency
and distribution very important from practical point of view (we live in the world
of multicore systems and cocurrency is omnipresent!!)
I have appended the AIX manpage for mmap. Maye it helps!
mmap or mmap64 Subroutine
Purpose
Maps a file-system object into virtual memory.
Library
Standard C library (libc.a)
Syntax
#include <sys/types.h>
#include <sys/mman.h>
void *mmap (addr, len, prot, flags, fildes, off)
void * addr;
size_t len;
int prot, flags, fildes;
off_t off;
void *mmap64 (addr, len, prot, flags, fildes, off)
void * addr;
size_t len;
int prot, flags, fildes;
off64_t off;
Description
Attention: A file-system object should not be simultaneously mapped using both the mmap
and shmat subroutines. Unexpected results may occur when references are made beyond the
end of the object.
The mmap subroutine creates a new mapped file or anonymous memory region by establishing
a mapping between a process-address space and a file-system object. Care needs to be
taken when using the mmap subroutine if the program attempts to map itself. If the page
containing executing instructions is currently referenced as data through an mmap
mapping, the program will hang. Use the -H4096 binder option, and that will put the
executable text on page boundaries. Then reset the file that contains the executable
material, and view via an mmap mapping.
A region created by the mmap subroutine cannot be used as the buffer for read or write
operations that involve a device. Similarly, an mmap region cannot be used as the buffer
for operations that require either a pin or xmattach operation on the buffer.
Modifications to a file-system object are seen consistently, whether accessed from a
mapped file region or from the read or write subroutine.
Child processes inherit all mapped regions from the parent process when the fork
subroutine is called. The child process also inherits the same sharing and protection
attributes for these mapped regions. A successful call to any exec subroutine will unmap
all mapped regions created with the mmap subroutine.
The mmap64 subroutine is identical to the mmap subroutine except that the starting
offset for the file mapping is specified as a 64-bit value. This permits file mappings
which start beyond OFF_MAX.
In the large file enabled programming environment, mmap is redefined to be mmap64.
If the application has requested SPEC1170 compliant behavior then the st_atime field of
the mapped file is marked for update upon successful completion of the mmap call.
If the application has requested SPEC1170 compliant behavior then the st_ctime and
st_mtime fields of a file that is mapped with MAP_SHARED and PROT_WRITE are marked for
update at the next call to msync subroutine or munmap subroutine if the file has been
modified.
Parameters
addr
Specifies the starting address of the memory region to be mapped. When the
MAP_FIXED flag is specified, this address must be a multiple of the page size
returned by the sysconf subroutine using the _SC_PAGE_SIZE value for the Name
parameter. A region is never placed at address zero, or at an address where it
would overlap an existing region.
len
Specifies the length, in bytes, of the memory region to be mapped. The system
performs mapping operations over whole pages only. If the len parameter is not a
multiple of the page size, the system will include in any mapping operation the
address range between the end of the region and the end of the page containing the
end of the region.
prot
Specifies the access permissions for the mapped region. The sys/mman.h file defines
the following access options:
PROT_READ
Region can be read.
PROT_WRITE
Region can be written.
PROT_EXEC
Region can be executed.
PROT_NONE
Region cannot be accessed.
The prot parameter can be the PROT_NONE flag, or any combination of the PROT_READ
flag, PROT_WRITE flag, and PROT_EXEC flag logically ORed together. If the PROT_NONE
flag is not specified, access permissions may be granted to the region in addition
to those explicitly requested. However, write access will not be granted unless the
PROT_WRITE flag is specified. Note: The operating system generates a SIGSEGV signal
if a program attempts an access that exceeds the access permission given to a
memory region. For example, if the PROT_WRITE flag is not specified and a program
attempts a write access, a SIGSEGV signal results.
If the region is a mapped file that was mapped with the MAP_SHARED flag, the mmap
subroutine grants read or execute access permission only if the file descriptor
used to map the file was opened for reading. It grants write access permission only
if the file descriptor was opened for writing.
If the region is a mapped file that was mapped with the MAP_PRIVATE flag, the mmap
subroutine grants read, write, or execute access permission only if the file
descriptor used to map the file was opened for reading. If the region is an
anonymous memory region, the mmap subroutine grants all requested access
permissions.
fildes
Specifies the file descriptor of the file-system object or of the shared memory
object to be mapped. If the MAP_ANONYMOUS flag is set, the fildes parameter must be
-1. After the successful completion of the mmap subroutine, the file or the shared
memory object specified by the fildes parameter can be closed without affecting the
mapped region or the contents of the mapped file. Each mapped region creates a file
reference, similar to an open file descriptor, which prevents the file data from
being deallocated. Note: The mmap subroutine supports the mapping of shared memory
object and regular files only. An mmap call that specifies a file descriptor for a
special file fails, returning the ENODEV error code. An example of a file
descriptor for a special file is one that might be used for mapping either I/O or
device memory.
off
Specifies the file byte offset at which the mapping starts. This offset must be a
multiple of the page size returned by the sysconf subroutine using the
_SC_PAGE_SIZE value for the Name parameter.
flags
Specifies attributes of the mapped region. Values for the flags parameter are
constructed by a bitwise-inclusive ORing of values from the following list of
symbolic names defined in the sys/mman.h file:
MAP_FILE
Specifies the creation of a new mapped file region by mapping the file
associated with the fildes file descriptor. The mapped region can extend
beyond the end of the file, both at the time when the mmap subroutine is
called and while the mapping persists. This situation could occur if a file
with no contents was created just before the call to the mmap subroutine, or
if a file was later truncated. However, references to whole pages following
the end of the file result in the delivery of a SIGBUS signal. Only one of
the MAP_FILE and MAP_ANONYMOUS flags must be specified with the mmap
subroutine.
MAP_ANONYMOUS
Specifies the creation of a new, anonymous memory region that is initialized
to all zeros. This memory region can be shared only with the descendants of
the current process. When using this flag, the fildes parameter must be -1.
Only one of the MAP_FILE and MAP_ANONYMOUS flags must be specified with the
mmap subroutine.
MAP_ VARIABLE
Specifies that the system select an address for the new memory region if the
new memory region cannot be mapped at the address specified by the addr
parameter, or if the addr parameter is null. Only one of the MAP_VARIABLE
and MAP_FIXED flags must be specified with the mmap subroutine.
MAP_FIXED
Specifies that the mapped region be placed exactly at the address specified
by the addr parameter. If the application has requested SPEC1170 complaint
behavior and the mmap request is successful, the mapping replaces any
previous mappings for the process' pages in the specified range. If the
application has not requested SPEC1170 compliant behavior and a previous
mapping exists in the range then the request fails. Only one of the
MAP_VARIABLE and MAP_FIXED flags must be specified with the mmap subroutine.
MAP_SHARED
When the MAP_SHARED flag is set, modifications to the mapped memory region
will be visible to other processes that have mapped the same region using
this flag. If the region is a mapped file region, modifications to the
region will be written to the file.
You can specify only one of the MAP_SHARED or MAP_PRIVATE flags with the
mmap subroutine. MAP_PRIVATE is the default setting when neither flag is
specified unless you request SPEC1170 compliant behavior. In this case, you
must choose either MAP_SHARED or MAP_PRIVATE.
MAP_PRIVATE
When the MAP_PRIVATE flag is specified, modifications to the mapped region
by the calling process are not visible to other processes that have mapped
the same region. If the region is a mapped file region, modifications to the
region are not written to the file.
If this flag is specified, the initial write reference to an object page
creates a private copy of that page and redirects the mapping to the copy.
Until then, modifications to the page by processes that have mapped the same
region with the MAP_SHARED flag are visible.
You can specify only one of the MAP_SHARED or MAP_PRIVATE flags with the
mmap subroutine. MAP_PRIVATE is the default setting when neither flag is
specified unless you request SPEC1170 compliant behavior. In this case, you
must choose either MAP_SHARED or MAP_PRIVATE.
Return Values
If successful, the mmap subroutine returns the address at which the mapping was placed.
Otherwise, it returns -1 and sets the errno global variable to indicate the error.
Error Codes
Under the following conditions, the mmap subroutine fails and sets the errno global
variable to:
EACCES
The file referred to by the fildes parameter is not open for read access, or the
file is not open for write access and the PROT_WRITE flag was specified for a
MAP_SHARED mapping operation. Or, the file to be mapped has enforced locking
enabled and the file is currently locked.
EAGAIN
The fildes parameter refers to a device that has already been mapped.
EBADF
The fildes parameter is not a valid file descriptor, or the MAP_ANONYMOUS flag was
set and the fildes parameter is not -1.
EFBIG
The mapping requested extends beyond the maximum file size associated with fildes.
EINVAL
The flags or prot parameter is invalid, or the addr parameter or off parameter is
not a multiple of the page size returned by the sysconf subroutine using the
_SC_PAGE_SIZE value for the Name parameter.
EINVAL
The application has requested SPEC1170 compliant behavior and the value of flags is
invalid (neither MAP_PRIVATE nor MAP_SHARED is set).
EMFILE
The application has requested SPEC1170 compliant behavior and the number of mapped
regions would excedd and implementation-dependent limit (per process or per
system).
ENODEV
The fildes parameter refers to an object that cannot be mapped, such as a terminal.
ENOMEM
There is not enough address space to map len bytes, or the application has not
requested Single UNIX Specification, Version 2 compliant behavior and the MAP_FIXED
flag was set and part of the address-space range (addr, addr+len) is already
allocated.
ENXIO
The addresses specified by the range (off, off+len) are invalid for the fildes
parameter.
EOVERFLOW
The mapping requested extends beyond the offset maximum for the file description
associated with fildes.
Related Information
The exec (exec: execl, execle, execlp, execv, execve, execvp, or exect Subroutine)
subroutine, fork (fork, f_fork, or vfork Subroutine) subroutine, munmap (munmap
Subroutine) subroutine, read subroutine, shm_open subroutine, shm_unlink subroutine,
shmat subroutine, sysconf subroutine, write subroutine.
The pin kernel service, xmattach kernel service.
List of Memory Manipulation Services, List of Memory Mapping Services, Understanding
Memory Mapping in AIX 5L Version 5.3 General Programming Concepts: Writing and Debugging
Programs.
I was just curious and performed the qa-cilk test for 32bit version of gcc and xlc.
Here are the results:
IBM Compiler xlc (32bit)
------------------------
# gmake
./build
Discovered AIX:
gmake[1]: Entering directory `/tmp/newlisp-10.0.3'
gmake -f makefile_aix_utf8_xlc
gmake[2]: Entering directory `/tmp/newlisp-10.0.3'
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX newlisp.c
729 1500-010: (W) WARNING in main: Infinite loop. Program may not stop.
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-symbol.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-math.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-list.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-liststr.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-string.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-filesys.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-sock.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-import.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-xml.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-web.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-matrix.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-debug.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX nl-utf8.c
xlc_r -c -g -O2 -DSUPPORT_UTF8 -DAIX pcre.c
xlc_r newlisp.o nl-symbol.o nl-math.o nl-list.o nl-liststr.o nl-string.o nl-filesys.o nl-sock.o nl-import.o nl-xml.o nl-web.o nl-matrix.o nl-debug.o nl-utf8.o pcre.o -lm -ldl -lrt -lnsl -o newlisp
gmake[2]: Leaving directory `/tmp/newlisp-10.0.3'
gmake[1]: Leaving directory `/tmp/newlisp-10.0.3'
# ./newlisp qa-cilk
(abort) -> true
(sync) -> ()
(sync) -> (4792460 4280556 1069190 4083820 5017818 4001856 4489308 1732826 4178154 4534368)
N of processes -> 10
waiting to sync -> true
fibo(7) with 41 processes -> 21
pid: 4280560 p1:400
pid: 4792464 p2:500
true
>>>>> THE CILK API TESTED SUCESSFULL
GNU C Compiler gcc (32bit)
--------------------------
# gmake
./build
Discovered AIX:
gmake[1]: Entering directory `/tmp/newlisp-10.0.3'
gmake -f makefile_aix_utf8_gcc
gmake[2]: Entering directory `/tmp/newlisp-10.0.3'
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX newlisp.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-symbol.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-math.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-list.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-liststr.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-string.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-filesys.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-sock.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-import.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-xml.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-web.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-matrix.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-debug.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-utf8.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX pcre.c
gcc newlisp.o nl-symbol.o nl-math.o nl-list.o nl-liststr.o nl-string.o nl-filesys.o nl-sock.o nl-import.o nl-xml.o nl-web.o nl-matrix.o nl-debug.o nl-utf8.o pcre.o -lm -ldl -lrt -lnsl -o newlisp
gmake[2]: Leaving directory `/tmp/newlisp-10.0.3'
gmake[1]: Leaving directory `/tmp/newlisp-10.0.3'
# ./newlisp qa-cilk
(abort) -> true
(sync) -> ()
(sync) -> (5210178 3985454 4087876 5427248 2719950 1069220 2875640 5550318 4284462 4792348)
N of processes -> 10
waiting to sync -> true
fibo(7) with 41 processes -> 21
pid: 5210182 p1:400
pid: 3985458 p2:500
true
>>>>> THE CILK API TESTED SUCESSFULL
Does it mean that Cilk API is OK?
The 64bit versions also return the same result.
Yes, as it turns out the Cilk API always worked on AIX, only 'share' was affected. As it turns out from your results using test-mmap.c, the mmmap() function was working Ok in all compile flavors. The problem was only with the 'share' function in different portions of the code.
After a fix, 'share' should now work on AIX in all flavors too.
I posted/emailed a new newlisp-10.0.3.tgz in/to the usual places.
Looks like we will have both compilers and 32-bit and 64-bit supported on AIX :-)
Congratulation!
It works now out of the box:-)
Here is the output of all runs:
# gmake
./build
Discovered AIX:
gmake[1]: Entering directory `/tmp/newlisp-10.0.3'
gmake -f makefile_aixLP64_utf8_xlc
gmake[2]: Entering directory `/tmp/newlisp-10.0.3'
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX newlisp.c
729 1500-010: (W) WARNING in main: Infinite loop. Program may not stop.
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-symbol.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-math.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-list.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-liststr.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-string.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-filesys.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-sock.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-import.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-xml.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-web.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-matrix.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-debug.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-utf8.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX pcre.c
OBJECT_MODE=64 xlc_r newlisp.o nl-symbol.o nl-math.o nl-list.o nl-liststr.o nl-string.o nl-filesys.o nl-sock.o nl-import.o nl-xml.o nl-web.o nl-matrix.o nl-debug.o nl-utf8.o pcre.o -lm -ldl -lrt -lnsl -o newlisp
gmake[2]: Leaving directory `/tmp/newlisp-10.0.3'
gmake[1]: Leaving directory `/tmp/newlisp-10.0.3'
# ./newlisp qa-dot
Testing built-in functions ...
Testing contexts as objects and scoping rules ...
total time: 5626
>>>>> TESTING: ./newlisp FINISHED WITH ERRORS:
UTF-8 upper-case: failed
UTF-8 lower-case: failed
# ./newlisp qa-cilk
(abort) -> true
(sync) -> ()
(sync) -> (5357676 1069250 4178148 5156980 2875452 4014164 5550282 5541938 2150542 4124856)
N of processes -> 10
waiting to sync -> true
fibo(7) with 41 processes -> 21
pid: 5357680 p1:400
pid: 1069254 p2:500
true
>>>>> THE CILK API TESTED SUCESSFULL
====================== xlc_r (64bit) =======================
# gmake aix64
gmake -f makefile_aixLP64_utf8_xlc
gmake[1]: Entering directory `/tmp/newlisp-10.0.3'
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX newlisp.c
729 1500-010: (W) WARNING in main: Infinite loop. Program may not stop.
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-symbol.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-math.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-list.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-liststr.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-string.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-filesys.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-sock.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-import.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-xml.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-web.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-matrix.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-debug.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-utf8.c
xlc_r -q64 -c -g -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX pcre.c
OBJECT_MODE=64 xlc_r newlisp.o nl-symbol.o nl-math.o nl-list.o nl-liststr.o nl-string.o nl-filesys.o nl-sock.o nl-import.o nl-xml.o nl-web.o nl-matrix.o nl-debug.o nl-utf8.o pcre.o -lm -ldl -lrt -lnsl -o newlisp
gmake[1]: Leaving directory `/tmp/newlisp-10.0.3'
# ./newlisp qa-dot
Testing built-in functions ...
Testing contexts as objects and scoping rules ...
total time: 5622
>>>>> TESTING: ./newlisp FINISHED WITH ERRORS:
UTF-8 upper-case: failed
UTF-8 lower-case: failed
# ./newlisp qa-cilk
(abort) -> true
(sync) -> ()
(sync) -> (4825322 4083744 2887814 4489406 4444260 5156966 761898 4329518 1732730 2363604)
N of processes -> 10
waiting to sync -> true
fibo(7) with 41 processes -> 21
pid: 4825326 p1:400
pid: 4083748 p2:500
true
>>>>> THE CILK API TESTED SUCESSFULL
======================= gcc (32bit) ========================
gmake aix_gcc
gmake -f makefile_aix_utf8_gcc
gmake[1]: Entering directory `/tmp/newlisp-10.0.3'
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX newlisp.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-symbol.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-math.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-list.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-liststr.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-string.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-filesys.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-sock.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-import.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-xml.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-web.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-matrix.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-debug.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX nl-utf8.c
gcc -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DSUPPORT_UTF8 -DAIX pcre.c
gcc newlisp.o nl-symbol.o nl-math.o nl-list.o nl-liststr.o nl-string.o nl-filesys.o nl-sock.o nl-import.o nl-xml.o nl-web.o nl-matrix.o nl-debug.o nl-utf8.o pcre.o -lm -ldl -lrt -lnsl -o newlisp
gmake[1]: Leaving directory `/tmp/newlisp-10.0.3'
# ./newlisp qa-dot
Testing built-in functions ...
Testing contexts as objects and scoping rules ...
total time: 5574
>>>>> TESTING: ./newlisp FINISHED WITH ERRORS:
UTF-8 upper-case: failed
UTF-8 lower-case: failed
# ./newlisp qa-cilk
(abort) -> true
(sync) -> ()
(sync) -> (5017634 4001928 4661376 5124250 4825114 4567256 1970358 4096020 4362456 4313180)
N of processes -> 10
waiting to sync -> true
fibo(7) with 41 processes -> 21
pid: 5017638 p1:400
pid: 4001932 p2:500
true
>>>>> THE CILK API TESTED SUCESSFULL
======================= gcc (64bit) ========================
# gmake aix64_gcc
gmake -f makefile_aixLP64_utf8_gcc
gmake[1]: Entering directory `/tmp/newlisp-10.0.3'
gcc -maix64 -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX newlisp.c
gcc -maix64 -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-symbol.c
gcc -maix64 -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-math.c
gcc -maix64 -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-list.c
gcc -maix64 -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-liststr.c
gcc -maix64 -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-string.c
gcc -maix64 -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-filesys.c
gcc -maix64 -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-sock.c
gcc -maix64 -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-import.c
gcc -maix64 -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-xml.c
gcc -maix64 -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-web.c
gcc -maix64 -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-matrix.c
gcc -maix64 -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-debug.c
gcc -maix64 -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX nl-utf8.c
gcc -maix64 -Wall -pedantic -fno-strict-aliasing -Wno-uninitialized -Wno-long-long -c -O2 -DNEWLISP64 -DSUPPORT_UTF8 -DAIX pcre.c
gcc newlisp.o nl-symbol.o nl-math.o nl-list.o nl-liststr.o nl-string.o nl-filesys.o nl-sock.o nl-import.o nl-xml.o nl-web.o nl-matrix.o nl-debug.o nl-utf8.o pcre.o -maix64 -lm -ldl -lrt -lnsl -o newlisp
gmake[1]: Leaving directory `/tmp/newlisp-10.0.3'
# ./newlisp qa-dot
Testing built-in functions ...
Testing contexts as objects and scoping rules ...
total time: 5640
>>>>> TESTING: ./newlisp FINISHED WITH ERRORS:
UTF-8 upper-case: failed
UTF-8 lower-case: failed
# ./newlisp qa-cilk
(abort) -> true
(sync) -> ()
(sync) -> (5541896 4124840 1069242 5357668 4313196 4325506 4284632 4087944 540726 4648974)
N of processes -> 10
waiting to sync -> true
fibo(7) with 41 processes -> 21
pid: 5541900 p1:400
pid: 4124844 p2:500
true
>>>>> THE CILK API TESTED SUCESSFULL
Now I am about to start the more comprehensive test:
./newlisp qa-dot
Testing built-in functions ...
Testing contexts as objects and scoping rules ...
total time: 5992
>>>>> TESTING: ./newlisp FINISHED WITH ERRORS:
UTF-8 upper-case: failed
UTF-8 lower-case: failed
./newlisp qa-dictionary
N associations: 10000
cells before creating Lex: 80770
aquire from association list: 18 ms
time to aquire one entry: 1.800 μs
time reading and verifying entries ... : 30 ms
time to read and verify one entry: 3.000 μs
time to only read one entry: 2.200 μs
cells after creating Lex: 120783
N symbols in Lex: 10000
save to file: 482 ms
time to delete namespace: 27 ms
time to delete one symbol: 2.700 μs
cells after deleting Lex: 110787
time to load namespace: 59 ms
cells after loading Lex: 120789
time to save Lex2: 436 ms
>>>>> DICTIONARY API TESTED SUCCESSFUL
./newlisp qa-xml
testing: xml-parse and xml-type-tags
true true true true
XML API OK
Testing xml-parse with callback feature
parsed expresson:(NAME "apple"), source:<NAME>apple</NAME>
parsed expresson:(COLOR "red"), source:<COLOR>red</COLOR>
parsed expresson:(PRICE "0.80"), source:<PRICE>0.80</PRICE>
parsed expresson:(NAME "orange"), source:<NAME>orange</NAME>
parsed expresson:(COLOR "orange"), source:<COLOR>orange</COLOR>
parsed expresson:(PRICE "1.00"), source:<PRICE>1.00</PRICE>
parsed expresson:(NAME "banana"), source:<NAME>banana</NAME>
parsed expresson:(COLOR "yellow"), source:<COLOR>yellow</COLOR>
parsed expresson:(PRICE "0.60"), source:<PRICE>0.60</PRICE>
>>>>> XML CALLBACK TESTED SUCCESSFUL
./newlisp qa-setsig
..sig-1 ..sig-2 ..sig-3 ..sig-4 ..sig-5 ..sig-6 ..sig-7 ..sig-8 ..sig-10 ..sig-11 ..sig-12 ..sig-13 ..sig-14 ..sig-15 ..sig-16
sent: (1 2 3 4 5 6 7 8 10 11 12 13 14 15 16)
received: (1 2 3 4 5 6 7 8 10 11 12 13 14 15 16)
>>>>> SIGNAL TESTING SUCCESSFUL
./newlisp qa-net
target URL: http://localhost:10001//tmp/newlisp-10.0.3/qa-junk.txt
host: localhost
port: 10001
path: /tmp/newlisp-10.0.3/qa-junk.txt
Server already running ...
waiting for server ...
net-eval poblem with normal mode ->ERROR
and it is hanging here:-(
Any idea?
Great news. The 'net-eval' fail is probably a timeout issue, it may go away by just repeating the test or executing "./newlisp net-eval" on its own.
Try out manually:
[1] 11283
~/newlisp-10.0.3> ./newlisp
newLISP v.10.0.3 on OSX IPv4 UTF-8, execute 'newlisp -h' for more info.
> (net-eval "127.0.0.1" 12345 "(+ 3 4)")
7
>
if this works then the timeouts are too small in 66 and 70 in qa-net
If this does not work try this:
[1] 11301
~/newlisp-10.0.3> telnet 127.0.0.1 12345
Trying 127.0.0.1...
Connected to localhost.
Escape character is '^]'.
(symbols) <-- after connecting enter any expression to evaluate
(! != $ $0 $1 $10 $11 $12 ... lots of output ... xml-type-tags zero? | ~)
This checks if server mode works at all using telnet.
... I am surprised we see:
waiting for server ...
We should see:
waiting for server ...
why does it say the server is already running? Perhaps something else on your machine using port 10001, which is the port qa-net uses for newLISP server mode. This makes the client process think there is newLISP already on the server side.
I'm sorry that I didn't pay due attention to "Server already running ..." message myself.
The port 10001 as SCP configuration port on AIX is already in use. I changed the port
to 12321 in qa-comma, qa-dot and qa-net files and started the qa-net test manually.
It works as expected:
[1] 5452018
# ./newlisp
newLISP v.10.0.3 64-bit on AIX IPv4 UTF-8, execute 'newlisp -h' for more info.
> (net-eval "127.0.0.1" 12321 "(+ 3 4)")
7
> (exit)
#
But
target URL: http://localhost:12321//tmp/newlisp-10.0.3/qa-junk.txt
host: localhost
port: 12321
path: /tmp/newlisp-10.0.3/qa-junk.txt
Starting server on localhost
waiting for server ...
net-eval normal mode ->OK
is hanging again! Any idea?
By the way what/where is net-eval.c file? Maybe you have meant qa-net?!
Yes, I meant qa-net (now corrected in the post).
The next test is the following (manually):
"nERR: invalid function : (abc)n"
>
and in qa-net line 75, the output is checked to be correct. This checks if the servermode can recover the network connection after going through and initialization and resettting routine. You could check this manually by connecting to the server with telnet (as shown earlier) and enter: (abc) and it should come back with a an error message, and another command, e.g. (symbols) should still work.
We basically would have to go through all of qa-net's routines manually to find out what the problem is. I have seen qa-net fail only in the places where (sleep ..) statements are inserted, and only on older slower machines, or when running in VM environments like Virtual Box.
There are also a examples/server and examples/client test programs in the source distribution. Start the server first without parameters, then the client with only: localhost as argument in another peocess. Lines entered in the client will be sent back uppercased by the server. The programs use the port 1111:
[1] 12998
~/newlisp-10.0.3/examples>
Server started
~/newlisp-10.0.3/examples> ./client localhost
Connected!
input or 'quit' to exit:sdkjfghsldkjgh
received:sdkjfghsldkjgh
SDKJFGHSLDKJGH
input or 'quit' to exit:
These two programs also do a test of basic networking functions. Practically all of what is in there is also tested in qa-dot already. So I assume, its something peculiar about server mode in qa-net.
Another basic test for server mode would be, to start the server, just like before, but in a directory containing some HTML files. You should then be able to put http://localhost:12345/thePage.html in a web browser and see the page.
The expression (net-eval "localhost" 12321 "(abc)") causes the problem.
Without a timeout value it will be hanging!! I treid it with different timeout values:
[1] 1069264
#./newlisp
newLISP v.10.0.3 64-bit on AIX IPv4 UTF-8, execute 'newlisp -h' for more info.
> (net-eval "localhost" 12321 {(+ 3 4)} 1000)
7
> (net-eval "localhost" 12321 "(abc)" 2000)
nil
> (net-eval "localhost" 12321 "(abc)" 5000)
nil
> (net-eval "localhost" 12321 "(abc)" 10000)
nil
>
It would be reasonable to add a timeout to the expression (net-eval host port "(abc)")
at line 75 in "qa-net" file:
+75 (if (= (net-eval host port "(abc)") "nERR: invalid function : (abc)n")
+76 (set 'result2 (println "net-eval error reporting ->OK"))
+77 (println "net-eval problem with error reporting ->ERROR"))
Any idea why we don't get back the expected error message?
I will also do the other tests that you have mentioned.
It would be interesting to know if the server still reacts to 'net-eval' with {(+ 3 4)} correctly after you have a 'nil' return.
Also after it returned with 'nil' do a: (net-error) and then a: (sys-error) to see error information from newLISP's own routines and perhaps from AIX when using: (sys-error).
Before that clear 'sys-error' to 0 by doing a: (sys-error 0), to be sure that the system error you get is not from something you did earlier.
When a 'net-xxx' routines returns 'nil' it is always an error condition, which can be checked for with 'net-error'. Sometimes 'sys-error' offers additional OS information.
The expression (net-eval "localhost" 12321 "(abc)") causes the problem.
Without a timeout value it will be hanging!! I tried it with different timeout values:
[1] 1069264
#./newlisp
newLISP v.10.0.3 64-bit on AIX IPv4 UTF-8, execute 'newlisp -h' for more info.
> (net-eval "localhost" 12321 {(+ 3 4)} 1000)
7
> (net-eval "localhost" 12321 "(abc)" 2000)
nil
> (net-eval "localhost" 12321 "(abc)" 5000)
nil
> (net-eval "localhost" 12321 "(abc)" 10000)
nil
>
It would be reasonable to add a timeout to the expression (net-eval host port "(abc)")
at line 75 in "qa-net" file:
+75 (if (= (net-eval host port "(abc)") "nERR: invalid function : (abc)n")
+76 (set 'result2 (println "net-eval error reporting ->OK"))
+77 (println "net-eval problem with error reporting ->ERROR"))
Any idea why we don't get back the expected error message?
I will also do the other tests that you have mentioned.
Quote
Thats the crucial point. A timeout doesn't help here. That's why I want you do do the other tests, confirming the timeout with: (net-error), looking for AIX system error messages and finding out if the server is still reacting after the timeout. I guess it will not be responsive. Is it down? Use the Unix utility ps to see what it is doing. At this moment I have no idea why it doesn't come back with an error message.
I have started newlisp as follows
And in another terminal done my tests
newLISP v.10.0.3 64-bit on AIX IPv4 UTF-8, execute 'newlisp -h' for more info.
> (net-eval "localhost" 12321 {(+ 3 4)} 1000)
7
> (net-eval "localhost" 12321 "(abc)" 2000)
nil
> (net-error)
(17 "ERR: Operation timed out")
> (sys-error)
(0 "Error 0")
> (net-eval "localhost" 12321 {(+ 3 4)} 1000)
7
> (net-eval "localhost" 12321 "(/ 1 0)" 1000)
nil
> (net-eval "localhost" 12321 {(+ 3 4)} 5000)
7
> (net-eval "localhost" 12321 "(abc)" 5000)
nil
> (net-eval "localhost" 12321 "(abc)" 50000)
nil
> (net-eval "localhost" 12321 "(* (* 4 5) (+ 30 45))" 1000)
1500
>
The content of log file:
Connected to 127.0.0.1 on Sat Mar 14 19:53:33 2009
[cmd]
(+ 3 4)(exit)
[/cmd]
7
Connected to 127.0.0.1 on Sat Mar 14 19:53:46 2009
[cmd]
(abc)(exit)
[/cmd]
Connected to 127.0.0.1 on Sat Mar 14 19:54:23 2009
[cmd]
(+ 3 4)(exit)
[/cmd]
7
Connected to 127.0.0.1 on Sat Mar 14 19:54:38 2009
[cmd]
(/ 1 0)(exit)
[/cmd]
Connected to 127.0.0.1 on Sat Mar 14 19:56:34 2009
[cmd]
(+ 3 4)(exit)
[/cmd]
7
Connected to 127.0.0.1 on Sat Mar 14 19:57:09 2009
[cmd]
(abc)(exit)
[/cmd]
Connected to 127.0.0.1 on Sat Mar 14 19:57:42 2009
[cmd]
(abc)(exit)
[/cmd]
Connected to 127.0.0.1 on Sat Mar 14 20:00:24 2009
[cmd]
(* (* 4 5) (+ 30 45))(exit)
[/cmd]
1500
And here is the output of server and client example:
Server side
-----------
# cd examples
# ../newlisp server
Server started
received:fischers fritz fischt frische fische
received:es hat geklappt!
received:quit
Client side
-----------
# cd examples
# ../newlisp client localhost
Connected!
input or 'quit' to exit:fischers fritz fischt frische fische
FISCHERS FRITZ FISCHT FRISCHE FISCHE
input or 'quit' to exit:es hat geklappt!
ES HAT GEKLAPPT!
input or 'quit' to exit:quit
bye bye!
input or 'quit' to exit:#
I have created the test.html file with the following content:
<head>Hello</head>
<body>
FISCHERS FRITZ FISCHT FRISCHE FISCHE
</body>
</html>
and opened the url http://myserver:12321/test.html in my browser
from another PC in another network and got
in browser window! Considering only these tests it seems that my newlisp version on AIX should be OK!
Unfortunately the problem that we observed above is still open:-(
I have also check system logs. No trace of any errors which could eventually be related to newlisp!
Show goes on!!
Thanks for all the testing, the picture is getting clearer now. Adding the sever log was also a good idea. If the last text is literaly, what you see in the browser:
then something is eating line-feeds between the server and the client and the browser. The HTTP header should not be visible in the browser and seems to be merged with the content. Only content between body tags should be visible. This could happen if line feeds are removed. What we should see in the browser is just:
Also the Content length with 86 is exactly the content-length of "test.html" with line-feeds removed. The header is written by newLISP server, that means that when it reads the file: test.html, it already got it without line-feeds. Try this in newLISP:
It should give the file size, in my case 92. The same number you see, should be the number in Content-length.
Also lets try the following:
[1] 14659
~/newlisp-10.0.3> telnet localhost 12321
Trying ::1...
telnet: connect to address ::1: Connection refused
Trying fe80::1...
telnet: connect to address fe80::1: Connection refused
Trying 127.0.0.1...
Connected to localhost.
Escape character is '^]'.
(abc)
ERR: invalid function : (abc)
(append "fischersrn" "fritzn")
"fischersrnfritzn"
Lets check both CR-LF and LF only.
As you, I believe that your AIX compiled version, is generally Ok, except for (1) server mode, where the error message/string disappears. (2) HTTP content arriving at the browser with line-feeds removed. Maybe generally loosing line feeds when reading files?
Probably both things are related! Does AIX use special line feed characters? What does a HEX dump of test.html look like?
Yes, I got exactly this output
in my browser window.
newLISP v.10.0.3 64-bit on AIX IPv4 UTF-8, execute 'newlisp -h' for more info.
> (define s (read-file "test.html"))
"<html>n<head>Hello</head>n<body>nFISCHERS FRITZ FISCHT FRISCHE FISCHEn</body>n</html>n"
> (length s)
86
>
Why do you mean that the result should be 92? Maybe because of "rn" instead of just "n"?
Telnet test leads to
[1] 4902916
# telnet localhost 12321
Trying...
Connected to localhost.
Escape character is '^]'.
(abc)
ERR: invalid function : (abc)
(append "fischersrn" "fritzn")
"fischersrnfritzn"
And here is the HEX dump of "test.html"
0x640x3e0x480x650x6c0x6c0x6f0x3c0x2f0x680x65
0x610x640x3e0x0a0x3c0x620x6f0x640x790x3e0x0a
0x460x490x530x430x480x450x520x530x200x460x52
0x490x540x5a0x200x460x490x530x430x480x540x20
0x460x520x490x530x430x480x450x200x460x490x53
0x430x480x450x0a0x3c0x2f0x620x6f0x640x790x3e
0x0a0x3c0x2f0x680x740x6d0x6c0x3e0x0a
As far as I know there is nothing special about LF on AIX. It is 'n' as on any other UNIX derivate.
I have done the same tests with our working version 10.0.3 on my Mac OSX (1.5.6) and everything worked very fine. No "problems" at all!
I'm a little bit confused at this moment. What goes here wrong?!
It is still more confusing if you look at the result of the following simple tests
which I have done with both versions of newlisp 10.0.3 compiled by xlc_r and gcc (the same for 32bit):
IBM Compile xlc_r
-----------------
# ./newlisp -L./test_run.log -c -d 12321 &
# ./newlisp
newLISP v.10.0.3 64-bit on AIX IPv4 UTF-8, execute 'newlisp -h' for more info.
> (net-eval "localhost" 12321 "(+ 2 3)" 1000)
nil
> (net-error)
nil
> (sys-error)
(0 "Error 0")
> (net-eval "localhost" 12321 "(abc)" 2000)
nil
> (net-error)
(17 "ERR: Operation timed out")
> (sys-error)
(0 "Error 0")
>
----------------------------------------------------------
# cat test_run.log
Connected to 127.0.0.1 on Sun Mar 15 18:15:38 2009
[cmd]
(+ 2 3)(exit)
[/cmd]
5
Connected to 127.0.0.1 on Sun Mar 15 18:16:03 2009
[cmd]
(abc)(exit)
[/cmd]
GNU C Compiler gcc
------------------
# ./newlisp -L./test_run.log -c -d 12321 &
# ./newlisp
newLISP v.10.0.3 64-bit on AIX IPv4 UTF-8, execute 'newlisp -h' for more info.
> (net-eval "localhost" 12321 "(+ 2 3)" 1000)
5
> (net-eval "localhost" 12321 "(abc)" 2000)
nil
>
----------------------------------------------------------
# cat test_run.log
Connected to 127.0.0.1 on Sun Mar 15 18:20:22 2009
[cmd]
(+ 2 3)(exit)
[/cmd]
5
Connected to 127.0.0.1 on Sun Mar 15 18:20:34 2009
[cmd]
(abc)(exit)
[/cmd]
Yes, 86 seems to be right. When I copy/pasted the test.html from the forum I got a trailing space after each line, so: 86 + 6 -> 92. But still, the browser should only show "Hello FISCHERS FRITZ FISCHT FRISCHE FISCHE". See your file served by Apache here: http://www.newlisp.org/test.html This is how it appears also on non-AIX systems when served with newLISP server. Somehow on the AIX machine the HTTP header gets merged with the message.
In the hex-dump and telnet session everything is Ok.
This is weird. With neither compiler we see the error message coming back, but with GCC at least we get the "5" back. In both cases the log shows that the "5" has been sent (error messages don't get logged currently, but I will add this).
I wonder if things change if server and client are on different computers?
To get further to the ground of this, one would have to watch the TCP traffic, compile with debug-statements etc.
I guess we have to leave stop our experiments here. If I get hold of an AIX machine one day, I can investigate further. I believe for non-server modes you should be fine with either compile of newLISP.
I have performed the same tests with windows XP as client and AIX as server.
The result is the same as before. I don't want to bother you further.
I know that it is very difficult to make a diagnosis remotely.
I will try to solve this puzzle myself if I have more time.
Thank you very much for your kind support!
I tried to build newlisp on AIX days ago. I managed to build it successfully after having cleaned up some macro statement in nl_filesystem.c and the result executable, newlisp, runs just fine in the machine where I built newlisp. However,when I copy the executable newlisp to another aix box and try to launch it, it just hung there and doesn't give me the prompt. After I press ctrl+c,it give me a option list with continue,abort,etc and I presss (c)ontinue, then the newlisp prompt just show up. Does any one meet such a problem and have any clue how to fix it?
What makefile are you using? There are two in the distribution, but I have never used them myself.
If you do a
You also should compile and run newlisp-x.x.x/util/types.c to check conformance of your AIX OS to either the LP64 or ILP32 memory model conventions. The expected results of running the
I've tried the makefiles for gcc and xlc but they both generated the same result. The executale newlisp works fine on the machine I built newlisp but hung on other aix machines.
For the changes to nl_filesys.c,I will give the details once i get access to the machine.
All aix machines are 64 bits and I built it with 64 bit makefile,I just can figure out why this would happen.
All right,I was back and got the access to the aix machine so I did some test based on lutz's suggestions.
I built the util/types.c and run it,following is the output of types.c:
type bytes
---------------
char 1
char * 4
void * 4
short int 2
int 4
long 4
long int 4
long long int 8
size_t 4
float 4
double 8
long double 8
wchar_t 2
size_t 4
off_t 4
time_t 4
format input output
------------------------------------------------
%d 0xffffffff -1
%u 0xffffffff 4294967295
%d 0x7fffffff 2147483647
%u 0x7fffffff 2147483647
%d 0x80000000 -2147483648
%u 0x80000000 2147483648
%lld 0x7fffffffffffffffLL 9223372036854775807
%llu 0x7fffffffffffffffLL 9223372036854775807
%llx 0xffffffffffffffffLL ffffffffffffffff
%llX 0x7fffffffffffffffLL 7fffffffffffffff
%llX 0x8000000000000000LL 8000000000000000
%llX 0xFFFFFFFFFFFFFFFFLL ffffffffffffffff
%llX 0x7FFFFFFFFFFFFFFFLL 7fffffffffffffff
%llX 0x8000000000000000LL 8000000000000000
%lld 0xffffffffffffffffLL -1
%llu 0xffffffffffffffffLL 18446744073709551615
%lld 0x7fffffffffffffffLL 9223372036854775807
%llu 0x7fffffffffffffffLL 9223372036854775807
%lld 0x8000000000000000LL -9223372036854775808
%llu 0x8000000000000000LL 9223372036854775808
CPU is big endian
The constant __BIG_ENDIAN__ is defined by the compiler
size of struct {char, short int, int} is: 8
size of struct {char, int, short int} is: 12
size of struct {char, short int, long, int} is: 12
size of struct {char, long, short int, int} is: 16
9223372036854775807 * 1000 = -1000
11 % -4 = 3
Following are the changes I'd made to nl_filesys.c:
--- newlisp-10.5.3.new/nl-filesys.c Wed Jul 31 14:06:39 2013
***************
*** 1615,1622 ****
--- 1615,1624 ----
#if defined(SUNOS) || defined(LINUX) || defined(CYGWIN)
memcpy(&thisFdSet, &myFdSet, sizeof(fd_set));
#else
+ #ifndef AIX
FD_COPY(&myFdSet, &thisFdSet);
#endif
+ #endif
if(mode == SELECT_READ_READY)
ready = select(FD_SETSIZE, &thisFdSet, NULL, NULL, &tv);
***************
*** 2584,2593 ****
--- 2586,2597 ----
struct tm *ltm;
#ifndef SUNOS
#ifndef OS2
+ #ifndef AIX
INT gmtoff;
UINT isdst;
#endif
#endif
+ #endif
#else /* WINDOWS */
TIME_ZONE_INFORMATION timeZone;
#endif
***************
*** 2607,2618 ****
--- 2611,2626 ----
ltm = localtime((time_t *)&tv.tv_sec);
#ifndef SUNOS
#ifndef OS2
+ #ifndef AIX
isdst = ltm->tm_isdst;
+ #endif
#ifdef CYGWIN
gmtoff = _timezone/60;
#else
+ #ifndef AIX
gmtoff = ltm->tm_gmtoff/60;
+ #endif
#endif
#endif
I still can't figure out the hung problem. If any one can give some suggestions?
Well,I did more tests and found that the built executable just runs find on some aix machines while hung on some aix machines. So looks like there're some execution restrictions on some machines. I'll go to our sys admin for this problem. Thanks
Glad you could figure it out Hugh. The output from types.c looks exactly as expected from the ILP32 memory model. I made a new
Also added rickyboy's improvements to
http://www.newlisp.org/downloads/development/inprogress/