./configure --with-stdc89 to test ANSI C conformity

[tiran]

BPO	4558
Nosy	@birkenfeld, @amauryfa, @vstinner, @tiran
Files	with_stdc89.patch with_stdc89_2.patch python-trunk-20081209-c89.patch: updated c89 patch for linux python3-20110520-c89.patch: update with_stdc89_3.patch with_stdc89_3.patch

^{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 2016-09-10.21:28:25.896>
created_at = <Date 2008-12-06.02:25:26.040>
labels = ['type-feature', 'build']
title = './configure --with-stdc89 to test ANSI C conformity'
updated_at = <Date 2016-09-10.21:28:25.895>
user = 'https://github.com/tiran'

bugs.python.org fields:

activity = <Date 2016-09-10.21:28:25.895>
actor = 'christian.heimes'
assignee = 'none'
closed = True
closed_date = <Date 2016-09-10.21:28:25.896>
closer = 'christian.heimes'
components = ['Build']
creation = <Date 2008-12-06.02:25:26.040>
creator = 'christian.heimes'
dependencies = []
files = ['12246', '12247', '12294', '22035', '30686', '30687']
hgrepos = []
issue_num = 4558
keywords = ['patch']
message_count = 16.0
messages = ['77097', '77102', '77106', '77141', '77144', '77146', '77148', '77149', '77156', '77240', '77369', '136338', '136413', '136438', '191765', '191766']
nosy_count = 5.0
nosy_names = ['georg.brandl', 'amaury.forgeotdarc', 'vstinner', 'christian.heimes', 'rpetrov']
pr_nums = []
priority = 'normal'
resolution = 'out of date'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'enhancement'
url = 'https://bugs.python.org/issue4558'
versions = ['Python 3.3', 'Python 3.4']

[tiran]

The patch adds a configure option --with-stdc89. It also fixes some edge
cases in socketmodule.h and linuxaudiodev so all related modules can be
build in c89 mode.

I'm planning to add the configure option to the build bots in order to
detect c89 incompatibilities like the usage of "inline" and // style
comments.

[birkenfeld]

Patch looks good.

[tiran]

The updated patch moves the compiler option to the right place.

[rpetrov]

There is no reason to overload configure script.

Please look into pyport.h :
#define Py_LOCAL(type) static type
#define Py_LOCAL_INLINE(type) static inline type

Someone add USE_INLINE - not finished.
I guess that other source files has to use macros from one
header(pyport.h) instead every C-file to resolve issue independently.

[tiran]

I think you don't get the point of the patch. The configure option
*enforces* the C89 standard. Every non C89 conform C feature like //
comments and the "inline" keyword will cause a compiler error.

The patch is meant for testing the Python code on the build bots.

[rpetrov]

Christian did you like to propose a separate configure flag for every
dialect that GCC support? I prefer variables like OPT.

[rpetrov]

P.S. About C++ comments ( // ) I think that another issue is already
opened (problem on AIX ? ) .

[tiran]

Python core code must be compliant with the C89 standard but we don't
have any checks for C89 compatibility. This patch adds a clean way to
check our code base.

You can verify my patch by adding a // comment somewhere in the code and
./configure --with-stdc89 && make.

[rpetrov]

make EXTRA_CFLAGS='-std=c89' works for me.

[rpetrov]

:-( the last my comment is incomplete : work for me after "minimal patch"

[rpetrov]

Socket module need to be patched too otherwise test_socket fail on UNIX
domain socket tests.

After clean build and with new patch the regression tests results are:
337 tests OK.
3 tests failed:
test_curses test_urllib2_localnet test_urllibnet
21 tests skipped:
test_aepack test_al test_applesingle test_bsddb185 test_cd test_cl
test_gl test_imgfile test_kqueue test_macos test_macostools
test_pep277 test_py3kwarn test_scriptpackages test_sqlite
test_startfile test_sunaudiodev test_unicode_file test_winreg
test_winsound test_zipfile64
Those skips are all expected on linux2.

Interesting for failed tests is that every succeed if run independently.

[rpetrov]

Issue with inline was resolved by configure macro.

[vstinner]

Is it still useful to ensure that ISO C89 compilers are supported in 2011?

[amauryfa]

Yes, vs2008 requires that variables must be declared at the start of a block, and IIRC there is a AIX compiler that does not allow // comments.

[tiran]

Updated patch for Python 3.4.

The feature is still useful. For example I found a bug in pymacro.h and fixed it in r84306.

[tiran]

Make that r6915dfddb3f6