Review Request 127834: ki18n: Fix theoretically possible use-after-free in gettext when using strict ANSI compilers

[Chusslove Illich]

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://git.reviewboard.kde.org/r/127834/#review95232
-----------------------------------------------------------


Ship it!




Ship It!

- Chusslove Illich


On Мај 5, 2016, 3:07 пре п., Michael Pyne wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://git.reviewboard.kde.org/r/127834/
> -----------------------------------------------------------
> 
> (Updated Мај 5, 2016, 3:07 пре п.)
> 
> 
> Review request for KDE Frameworks, Localization and Translation (l10n) and Chusslove Illich.
> 
> 
> Repository: ki18n
> 
> 
> Description
> -------
> 
> Coverity noted that some of code for message catalog lookup uses some pointer values after they are freed. Even though the use in question is a simple equality comparison against a different (valid) pointer, it is still undefined behavior according to the C (and C++) language specs and is therefore liable to cause miscompilation and who knows what other kinds of problems.
> 
> This code is not normally enabled, normally a code path that supports variable-length arrays is active, which is not susceptible to this bug.
> 
> Since VLAs are not supported even in current C++ versions, making VLA support mandatory is not feasible, so instead I opted to move the pointer comparisons to a point in the code where the comparison is valid, and then use the saved result later.
> 
> I have also reported the bug to GNU Gettext, since upstream still has the error. It is GNU Gettext bug 47847.
> 
> 
> Diffs
> -----
> 
>   src/gettext.h b06fc90 
> 
> Diff: https://git.reviewboard.kde.org/r/127834/diff/
> 
> 
> Testing
> -------
> 
> Code compiles and KDE applications still seem to work fine. I also tested with the changed code path forcibly enabled by disabling VLA support, and things still seemed to work.
> 
> There's not a lot of autotests to choose from, but the KLocalizedString test still passed.
> 
> 
> Thanks,
> 
> Michael Pyne
> 
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.kde.org/pipermail/kde-frameworks-devel/attachments/20160506/d6278342/attachment.html>