https://bugs.kde.org/show_bug.cgi?id=345307
Philippe Waroquiers <philippe.waroqui...@skynet.be> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |philippe.waroquiers@skynet.
| |be
--- Comment #32 from Philippe Waroquiers <philippe.waroqui...@skynet.be> ---
(In reply to Aleksandar Rikalo from comment #31)
> Created attachment 107894 [details]
> leak_cpp_interior additional .exps
>
> Re-based patch.
>
> Can someone take a look ?
As I understand, the problem is 'cleanly' fixed in a newer version of glibc, by
having valgrind calling __gnu_cxx::__freeres, but which is only present in
newer library version.
Having various different .exp files for the same test is difficult/painful to
maintain : a functional change implies to update all the templates (and so
implies to build and run valgrind in all the different setups for these
templates). And a real test failure is more difficult to analyse, because you
get a diff between the test output and all possible .exp candidates.
So, IMO, we better do not add new .exp files, in particular for a problem that
will be cleanly fixed by an upgrade to the library having the __freeres symbol.
The workaround is either to just ignore the test failure, or add a suppression
entry (see comment 5).
If we want a fix in the valgrind release, IMO, we better add the supp entry in
valgrind, as this should make both the test work, and suppress this for all the
users launched executables using the same lib version
--
You are receiving this mail because:
You are watching all bug changes.
