New warning on main: writing 1 byte into a region of size 0 [-Wstringop-overflow=]

[sobolevn]

Bug report

Bug description:

This happens in Modules/_decimal/libmpdec/io.c
Here:

cpython/Modules/_decimal/libmpdec/io.c

Lines 348 to 349 in a472244

	#define EXTRACT_DIGIT(s, x, d, dot) \
	if (s == dot) *s++ = '.'; *s++ = '0' + (char)(x / d); x %= d

CPython versions tested on:

CPython main branch

Operating systems tested on:

Linux

Linked PRs

• gh-129467: turn off false-positive -Wstringop-overflow in word_to_string() #116346

[skirpichev]

I doubt it's a new one, see #108562. See patch: #116346.

Probably someone (me?:)) should just post that to the mpdecimal maillist.

[sobolevn]

In any case, it affects our CI, so we should take appropriate steps to resolve this issue.

[skirpichev]

Ok, #116346 restored.

[vstinner]

Probably someone (me?:)) should just post that to the mpdecimal maillist.

Yes, otherwise, we will loose the patch next time that we update mpdecimal.

Note: we plan to remove the vendored libmpdecimal copy in Python 3.15: https://docs.python.org/dev/whatsnew/3.14.html#id11

[skirpichev]

Yes

Stefan participated in the old issue thread. So, he is aware on the problem. If this patch wasn't included in 4.0.0 - maybe he consider this as a non-issue for the mpdecimal.

next time that we update mpdecimal

I don't think there is much sense to do this. The mpdecimal seems to be a complete project (implementation of the IBM spec). For bugfixes or performance improvements people should just use upstream sources.

Note: we plan to remove the vendored libmpdecimal copy in Python 3.15

(Someone should convince Debian people to restore mpdecimal package.)

[vstinner]

Fixed by the change 6c63afc.