[issue2552] test_ctypes failed Python 2.6a2 Solaris 10 SUN C
Mark Lawrence added the comment: Can we close this as I cannot see how a test failure against a 6.5 year old alpha build can still be relevant? -- ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue2552 ___ ___ Python-bugs-list mailing list Unsubscribe: https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue2552] test_ctypes failed Python 2.6a2 Solaris 10 SUN C
Changes by Gregory P. Smith g...@krypto.org: -- resolution: - out of date status: languishing - closed ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue2552 ___ ___ Python-bugs-list mailing list Unsubscribe: https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue2552] test_ctypes failed Python 2.6a2 Solaris 10 SUN C
Changes by Stefan Krah stefan-use...@bytereef.org: -- nosy: -skrah ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue2552 ___ ___ Python-bugs-list mailing list Unsubscribe: https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue2552] test_ctypes failed Python 2.6a2 Solaris 10 SUN C
Mark Lawrence added the comment: Given the title is this still relevant given that Oracle Solaris 11 is closed source? -- nosy: +BreamoreBoy ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue2552 ___ ___ Python-bugs-list mailing list Unsubscribe: https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue2552] test_ctypes failed Python 2.6a2 Solaris 10 SUN C
Changes by Christian Heimes li...@cheimes.de: -- status: open - languishing ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue2552 ___ ___ Python-bugs-list mailing list Unsubscribe: https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue2552] test_ctypes failed Python 2.6a2 Solaris 10 SUN C
Changes by Terry J. Reedy tjre...@udel.edu: -- versions: +Python 2.7 -Python 2.6 ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue2552 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue2552] test_ctypes failed Python 2.6a2 Solaris 10 SUN C
Changes by Stefan Krah stefan-use...@bytereef.org: -- nosy: +skrah ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue2552 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue2552] test_ctypes failed Python 2.6a2 Solaris 10 SUN C
Changes by Jesús Cea Avión j...@jcea.es: -- nosy: +jcea ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue2552 ___ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue2552] test_ctypes failed Python 2.6a2 Solaris 10 SUN C
Martin v. Löwis [EMAIL PROTECTED] added the comment: If option --with-system-ffi is not specified and if the compiler is not gcc, raise an error. Up to here, it all sounded fine. What do you mean by error? Abort? That should not happen. Instead, setup.py skips modules if it doesn't find the proper prerequisites - which here is the pre-existing FFI implementation. __ Tracker [EMAIL PROTECTED] http://bugs.python.org/issue2552 __ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue2552] test_ctypes failed Python 2.6a2 Solaris 10 SUN C
Jean Brouwers [EMAIL PROTECTED] added the comment: Agreed. In addition, gcc -v writes to stderr and catching the output requires redirecting stderr or using os.popen3 or -4. The latter fail since the select module (needed by subprocess) is still missing at that time. Going back to Thomas' comments and before getting into implementation details, the best way to address the issue might be the following. If the compiler is not gcc then config option --with-system-ffi *must* be specified (and a libffi library must exist). If option --with-system-ffi is not specified and if the compiler is not gcc, raise an error. __ Tracker [EMAIL PROTECTED] http://bugs.python.org/issue2552 __ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue2552] test_ctypes failed Python 2.6a2 Solaris 10 SUN C
Jean Brouwers [EMAIL PROTECTED] added the comment: Instead of ... raise an error, it should read ... report the _ctypes module as missing, do not run any tests, etc. __ Tracker [EMAIL PROTECTED] http://bugs.python.org/issue2552 __ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue2552] test_ctypes failed Python 2.6a2 Solaris 10 SUN C
Jean Brouwers [EMAIL PROTECTED] added the comment: Yes, but the _ctypes module requires gcc, as Thomas Heller mentioned. The only thing which can be done is avoid building the _ctypes module if the compiler is not gcc. I did look at fixing that but have not found the proper way to do that. __ Tracker [EMAIL PROTECTED] http://bugs.python.org/issue2552 __ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue2552] test_ctypes failed Python 2.6a2 Solaris 10 SUN C
Martin v. Löwis [EMAIL PROTECTED] added the comment: Yes, but the _ctypes module requires gcc, as Thomas Heller mentioned. So the proper fix for that would be to port it to SunPro, of course. The only thing which can be done is avoid building the _ctypes module if the compiler is not gcc. I'm not sure whether that would be much better than the status quo, though. People would continue to complain that the ctypes module is not built. Why do you consider the current behaviour incorrect? I did look at fixing that but have not found the proper way to do that. There are several ways. One is to skip compilation in setup.py, for which you should then determine whether the compiler is gcc. For that, you could look at ac_compiler_gnu in configure.in, and set a variable in Makefile.pre.in if the compiler is GCC. Then, in setup.py, you could look for that variable as parsed from the Makefile, and skip _ctypes accordingly. Alternatively, you could check for __GNUC__ when compiling ctypes, and make the compiler produce a better-understandable #error. Regards, Martin __ Tracker [EMAIL PROTECTED] http://bugs.python.org/issue2552 __ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue2552] test_ctypes failed Python 2.6a2 Solaris 10 SUN C
Jean Brouwers [EMAIL PROTECTED] added the comment: For ctypes to compile with the SUN's C compiler, there is quite a bit of tricky asm code which needs to be ported (and tested). Most of that is in libffi and that library is only supported with gcc, it seems. But that is not mentioned explicitly anywhere either, see http://sources.redhat.com/libffi/. In any case, the _ctypes module does not compile with SUN C in the current situation. And it is not obvious why. Making that clear with some message that ctypes (rather _ctypes) requires gcc would help and might be sufficient. However, leaving the _ctypes module out altogether would be better. Just like other modules which have environment-specific dependencies, for example _tkinter. I'll take a look how that is handled. /Jean __ Tracker [EMAIL PROTECTED] http://bugs.python.org/issue2552 __ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue2552] test_ctypes failed Python 2.6a2 Solaris 10 SUN C
Jean Brouwers [EMAIL PROTECTED] added the comment: The setup.py module contains method detect_ctypes on line 1546 which handles the ctypes module building and testing. Would it be OK to not do any of that on Solaris, if the compiler is not gcc? Also, SUN Studio 12, the latest edition of the SUN C compiler has ... increased GCC __asm__ and __attribute__ compatibility ..., see http://blogs.sun.com/tatkar/entry/sunstudio_12_is_released and http://developers.sun.com/sunstudio/documentation/ss12/whatsnew.html. Maybe that one could compile libffi as is, for x86[_64]? __ Tracker [EMAIL PROTECTED] http://bugs.python.org/issue2552 __ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue2552] test_ctypes failed Python 2.6a2 Solaris 10 SUN C
Martin v. Löwis [EMAIL PROTECTED] added the comment: Would it be OK to not do any of that on Solaris, if the compiler is not gcc? That would be fine. __ Tracker [EMAIL PROTECTED] http://bugs.python.org/issue2552 __ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue2552] test_ctypes failed Python 2.6a2 Solaris 10 SUN C
Jean Brouwers [EMAIL PROTECTED] added the comment: Attached is one attempt to avoid building the _ctypes module if the compiler is not gcc: a patch for Python 2.6a2 with just 2 changes for setup.py based on the value of config variable CC. -- keywords: +patch Added file: http://bugs.python.org/file9954/setup.py.diff __ Tracker [EMAIL PROTECTED] http://bugs.python.org/issue2552 __ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue2552] test_ctypes failed Python 2.6a2 Solaris 10 SUN C
Martin v. Löwis [EMAIL PROTECTED] added the comment: I think this is incorrect; CC might also be gcc -pipe or /usr/local/bin/gcc, or some such, and would still be gcc (perhaps even /usr/local/bin/cc). __ Tracker [EMAIL PROTECTED] http://bugs.python.org/issue2552 __ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue2552] test_ctypes failed Python 2.6a2 Solaris 10 SUN C
Thomas Heller [EMAIL PROTECTED] added the comment: I'm curious: Why do you insist on using the sun compiler? Are there political or technical reasons? Another thing: in principle it should be possible to build a libffi shared library with gcc, and use it in a sun-c compiled python build. At least this would avoid the need to port assembler code from gcc to sun c. Python should be configured with './configure --with-systemffi' for that. -- nosy: +theller __ Tracker [EMAIL PROTECTED] http://bugs.python.org/issue2552 __ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue2552] test_ctypes failed Python 2.6a2 Solaris 10 SUN C
Gregory P. Smith [EMAIL PROTECTED] added the comment:
agreed, do not check to see if the string is == 'gcc'. Instead,
actually execute the CC compiler with a -v flag and test if 'gcc
version' in output.
something like this (but split into a few statements for readability):
if 'gcc version' not in os.popen(sysconfig.get_config_vars().get('CC',
'gcc') + ' -v').read():
# gcc is required for ctypes
--
nosy: +gregory.p.smith
__
Tracker [EMAIL PROTECTED]
http://bugs.python.org/issue2552
__
___
Python-bugs-list mailing list
Unsubscribe:
http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue2552] test_ctypes failed Python 2.6a2 Solaris 10 SUN C
Jean Brouwers [EMAIL PROTECTED] added the comment: Well, so much for the first attempt. I'll see what it takes to use something else, like ac_compiler_gnu. Unless Gregory's suggestion qualifies. Also, I am using SUN's compilers on Solaris 10 for 'political' and some technical reasons. And I am building alpha and beta Python releases mainly to help Python move forward. __ Tracker [EMAIL PROTECTED] http://bugs.python.org/issue2552 __ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue2552] test_ctypes failed Python 2.6a2 Solaris 10 SUN C
Martin v. Löwis [EMAIL PROTECTED] added the comment: Well, so much for the first attempt. I'll see what it takes to use something else, like ac_compiler_gnu. Unless Gregory's suggestion qualifies. I would prefer to see the configure test for gcc be recorded and passed through, over trying to detect gcc from its -v output, yes. Who knows what other compilers will do when passed the -v option? __ Tracker [EMAIL PROTECTED] http://bugs.python.org/issue2552 __ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue2552] test_ctypes failed Python 2.6a2 Solaris 10 SUN C
New submission from Jean Brouwers [EMAIL PROTECTED]: The ctypes test failed since the _ctypes module did not compile for 32- bit Python 2.6a2 on Solaris 10 (Opteron) built with the SUN C compiler. ... cc -KPIC -xtarget=native -DNDEBUG -xO5 -I. -I/.../tools/Python- 2.6a2/./Include -Ibuild/temp.solaris-2.10-i86pc-2.6/libffi/include - Ibuild/temp.solaris-2.10-i86pc-2.6/libffi -I/.../tools/Python- 2.6a2/Modules/_ctypes/libffi/src -I. -IInclude -I./Include - I/.../tools/Python-2.6a2/Include -I/.../tools/Python-2.6a2 -c .../tools/Python-2.6a2/Modules/_ctypes/_ctypes.c -o build/temp.solaris- 2.10-i86pc-2.6/.../tools/Python-2.6a2/Modules/_ctypes/_ctypes.o build/temp.solaris-2.10-i86pc-2.6/libffi/include/ffitarget.h, line 67: undefined symbol: FFI_DEFAULT_ABI build/temp.solaris-2.10-i86pc-2.6/libffi/include/ffitarget.h, line 68: non-constant enumerator value build/temp.solaris-2.10-i86pc-2.6/libffi/include/ffi.h, line 255: syntax error before or at: __attribute__ build/temp.solaris-2.10-i86pc-2.6/libffi/include/ffi.h, line 255: warning: old-style declaration or incorrect type for: __attribute__ build/temp.solaris-2.10-i86pc-2.6/libffi/include/ffi.h, line 255: warning: syntax error: empty declaration .../tools/Python-2.6a2/Modules/_ctypes/_ctypes.c, line 155: cannot recover from previous errors cc: acomp failed for /.../tools/Python-2.6a2/Modules/_ctypes/_ctypes.c ... -- components: Tests messages: 64941 nosy: MrJean1 severity: normal status: open title: test_ctypes failed Python 2.6a2 Solaris 10 SUN C versions: Python 2.6 __ Tracker [EMAIL PROTECTED] http://bugs.python.org/issue2552 __ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue2552] test_ctypes failed Python 2.6a2 Solaris 10 SUN C
Martin v. Löwis [EMAIL PROTECTED] added the comment: Are you interested in working on a patch? -- nosy: +loewis __ Tracker [EMAIL PROTECTED] http://bugs.python.org/issue2552 __ ___ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
