Python.h contains intermingled declarations

[encukou]

BPO	37191
Nosy	@vstinner, @encukou, @ambv, @jdemeyer, @lazka, @miss-islington
PRs	[3.8] bpo-37191: Avoid declaration-after-statement in header included from Python.h #13887 bpo-37191: Move PEP 590 vectorcall tests #13892 [3.8] bpo-37191: Move TestPEP590 from test_capi to test_call (GH-13892) #13897

^{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 2019-06-13.01:05:17.723>
created_at = <Date 2019-06-07.09:13:57.171>
labels = ['3.8', '3.9']
title = 'Python.h contains intermingled declarations'
updated_at = <Date 2019-06-13.01:05:17.717>
user = 'https://github.com/encukou'

bugs.python.org fields:

activity = <Date 2019-06-13.01:05:17.717>
actor = 'vstinner'
assignee = 'none'
closed = True
closed_date = <Date 2019-06-13.01:05:17.723>
closer = 'vstinner'
components = []
creation = <Date 2019-06-07.09:13:57.171>
creator = 'petr.viktorin'
dependencies = []
files = []
hgrepos = []
issue_num = 37191
keywords = ['patch']
message_count = 11.0
messages = ['344912', '344914', '344928', '344931', '344932', '344951', '344955', '344957', '344959', '344965', '345442']
nosy_count = 6.0
nosy_names = ['vstinner', 'petr.viktorin', 'lukasz.langa', 'jdemeyer', 'lazka', 'miss-islington']
pr_nums = ['13887', '13892', '13897']
priority = 'normal'
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = None
url = 'https://bugs.python.org/issue37191'
versions = ['Python 3.8', 'Python 3.9']

[encukou]

When compiled with GCC's -Werror=declaration-after-statement ("intermingled declarations" in PEP-7), cpython/abstract.h (included from <Python.h>) errors on vectorcall helper functions added in 3.8.0 Beta 1.

It's well within our rights to ignore this: since 3.6 we require intermingled declarations.
But, when re-compiling Fedora we've seen several projects fail with this warning (so far: pygobject3, python-dbus, xen; more will likely come).

Dear Release Manager, should we patch 3.8 to avoid this? The patch is simple, and it would give projects that we(re) dutifully tested with the Alphas one more release to adapt.

I don't think it's worth changing for 3.9 (but if we do we should test it).

[encukou]

pygobject3, python-dbus, xen

Correction:

• The Fedora packages failed to build; this might or might not be due to the projects themseves
• pygobject3 actually only warns on this

Anyway, we can usually take care of open-source stuff. Only take it as a litmus test for codebases we don't know about.

[vstinner]

PR 13887 is fine but I would prefer to still be able to use C99 style for variable declarations. More generally, I don't think that variable declaration is the only issue: static inline function is a new shiny thing which can cause other issues with some C compilers. So I propose a different approach, bpo-37194: "Move new vector private declarations to the internal C API".

[jdemeyer]

Should we revert these inline functions back to macros?

[jdemeyer]

It's well within our rights to ignore this: since 3.6 we require intermingled declarations.

It's not clear from PEP-7 if we require intermingled declarations only when compiling CPython itself, or also for external packages that include CPython headers. It would be prudent to stay C89-compatible for non-internal header files.

[lazka]

I've removed -Wdeclaration-after-statement in upstream pygobject: https://gitlab.gnome.org/GNOME/pygobject/merge_requests/119

[vstinner]

New changeset 740a84d by Victor Stinner in branch 'master':
bpo-37191: Move TestPEP590 from test_capi to test_call (GH-13892)
740a84d

[vstinner]

I've removed -Wdeclaration-after-statement in upstream pygobject: https://gitlab.gnome.org/GNOME/pygobject/merge_requests/119

Thanks! Enjoy C99! :-) IMHO C99 with variables in the middle of a function helps to write more reliable code: we better control the scope of a variable. It reduces the risk of using an uninitialized variable.

[miss-islington]

New changeset 5effd10 by Miss Islington (bot) in branch '3.8':
bpo-37191: Move TestPEP590 from test_capi to test_call (GH-13892)
5effd10

[vstinner]

New changeset 9689f80 by Victor Stinner (Petr Viktorin) in branch '3.8':
bpo-37191: Avoid declaration-after-statement in header included from Python.h (GH-13887)
9689f80

[vstinner]

Petr Viktorin fixed the issue, thanks ;-)