Repository navigation
update to autoconf2.65 #52756
Description
| BPO | 8510 |
|---|---|
| Nosy | @loewis, @doko42, @mdickinson, @vstinner |
| Files |
Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.
Show more details
GitHub fields:
assignee = None
closed_at = <Date 2010-05-11.10:34:12.856>
created_at = <Date 2010-04-23.15:26:17.423>
labels = ['build']
title = 'update to autoconf2.65'
updated_at = <Date 2010-05-11.10:34:12.843>
user = 'https://github.com/doko42'bugs.python.org fields:
activity = <Date 2010-05-11.10:34:12.843>
actor = 'mark.dickinson'
assignee = 'none'
closed = True
closed_date = <Date 2010-05-11.10:34:12.856>
closer = 'mark.dickinson'
components = ['Build']
creation = <Date 2010-04-23.15:26:17.423>
creator = 'doko'
dependencies = []
files = ['17055', '17070']
hgrepos = []
issue_num = 8510
keywords = ['patch', 'needs review']
message_count = 20.0
messages = ['104018', '104021', '104047', '104109', '104152', '104157', '104163', '104167', '104168', '104467', '104501', '105271', '105277', '105375', '105427', '105488', '105491', '105493', '105495', '105496']
nosy_count = 6.0
nosy_names = ['loewis', 'doko', 'mark.dickinson', 'vstinner', 'rpetrov', 'Arfrever']
pr_nums = []
priority = 'normal'
resolution = 'accepted'
stage = 'resolved'
status = 'closed'
superseder = None
type = None
url = 'https://bugs.python.org/issue8510'
versions = ['Python 2.7', 'Python 3.2']
Activity
doko42 commentedon Apr 23, 2010
update to autoconf2.65
doko42 commentedon Apr 23, 2010
see http://mail.python.org/pipermail/python-dev/2010-April/099639.html
loewis commentedon Apr 23, 2010
I don't think Python 2.7 should upgrade to a newer autoconf version at this point. For 3.2, we could try it out.
doko42 commentedon Apr 24, 2010
here is the diff for configure.in. The __EXTENSIONS__ macro is defined by 2.65 for Solaris, removed this explicit definition. running a test build
doko42 commentedon Apr 25, 2010
r80474: Replace AC_HELP_STRING with AS_HELP_STRING
doko42 commentedon Apr 25, 2010
r80475: s/AC_AIX/AC_USE_SYSTEM_EXTENSIONS/
doko42 commentedon Apr 25, 2010
r80478 converts all obsolete AC_TRY_* macros to AC_*_IFELSE, the outcome is ... nothing (but whitespace changes):
===================================================================
running configure with the old and the new version shows no diffs (besides config.log).
doko42 commentedon Apr 25, 2010
r80481: configure.in: Avoid autoconf warning: Assume C89 semantics that
RETSIGTYPE is always void (issue bpo-8510).
Keep the definition, although the python code itself doesn't use it anymore.
doko42 commentedon Apr 25, 2010
r80483: Makefile.pre.in (autoconf): Call autoconf/autoheader with -Wall to help the configure script to stay warning free.
rpetrov commentedon Apr 28, 2010
You don't need to add source in case of
AC_COMPILE_IFELSE([AC_LANG_SOURCE([[]])] ..
when the the test is ..main() { return 0 ;} ...
, posted long time ago as part of bpo-3754 ( minimal cross-compilation support for configure ).
Also thanks for fixing extra comma in after replacement of AC_TRY_COMPILE .
This part is save to be added to 2.7 build.
doko42 commentedon Apr 29, 2010
is there a version of the cross configure patch for 3.x?
doko42 commentedon May 8, 2010
the update to 2.65 on the py3k branch only did show whitespace changes, applying these changes to the trunk.
r80964: require autoconf-2.65
doko42 commentedon May 8, 2010
r80965: Replace AC_HELP_STRING with AS_HELP_STRING
r80966: s/AC_AIX/AC_USE_SYSTEM_EXTENSIONS/
r80969: convert all obsolete AC_TRY_* macros to AC_*_IFELSE
r80970: Avoid autoconf warning: Assume C89 semantics that RETSIGTYPE is always void
mdickinson commentedon May 9, 2010
r81004: Remove extra closing bracket and comma introduced in r80969. (This was causing misdetection of the OS X 10.5 SDK on Linux and OS X, and a test_platform failure on OS X.)
rpetrov commentedon May 10, 2010
I don't have access to source so my comment will be based only on diffs from above mentioned revisions.
r80969:
AC_LANG_PROGRAM ... with main function in body . It is legal as I don't know C compiler that fail to compile code like ...int main() { int main() {...} ...} but is better to left body empty. Counted three times. All those cases tests for compiler flags and I think that AC_LANG_SOURCE is enough.
body include #include "confdefs.h" - may be buggy (before) and now AC_LANG_PROGRAM will add all defines. Counted any times. I think that is save to remove #include "confdefs.h" from all those test cases.
mdickinson commentedon May 11, 2010
Roumen: good catch! Indeed, nested functions aren't legal in standard C, and the test for -fno-strict-aliasing now fails on my OS X 10.6.3 machine, where it used to pass:
checking whether gcc accepts -fno-strict-aliasing... no
config.log contains:
configure:5400: checking whether gcc accepts -fno-strict-aliasing
configure:5418: gcc -fno-strict-aliasing -c -g -O2 conftest.c >&5
conftest.c: In function 'main':
conftest.c:27: error: nested functions are disabled, use -fnested-functions to re-enable
configure:5418: $? = 1
configure: failed program was:
...
Fixed these cases above and removed the extra '#include "confdefs.h"' lines (which seem harmless, but unnecessary) in r81077 (trunk).
mdickinson commentedon May 11, 2010
Merged to py3k, and fixed up an additional nested 'int main()', in r81078.
mdickinson commentedon May 11, 2010
BTW, it looks as though the nested functions were introduced in r76030; nothing to do with Matthias's autoconf 2.65 update.
mdickinson commentedon May 11, 2010
Just double checked that r80969 didn't introduce any other non-whitespace changes: apart from the now-fixed OS X 10.5 SDK issue), the only other non-whitespace change was:
which looks harmless to me.
mdickinson commentedon May 11, 2010
Closing. Roumen, please open a new issue for any other configure issues that aren't related to this autoconf update.