[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Apps linked with old libstdc++ dlopening modules linked with the new one



Hello,

<short summary>
Applications linked with libstdc++5 and using GTK segfault when using
SCIM[1] as input method, and neither GTK nor SCIM nor the apps seem to
be to blame.
</short summary>

Since various days I've been experiencing weird and annoying crashes on
openoffice and epiphany.  I was kind of ignoring them, but now I need to
prepare some slides for a presentation and I can't ignore them anymore.

I checked on the bts, and noone seems to have experienced the same.

I checked the deps looking for something strange and related, and all
the crashing apps Depend on libstdc++5.  I checked among other packages
depending on libstdc++5 and I found I had drgeo installed, which crashed
as well.  Other non-GTK packages linking libstdc++5 (such as rig or
guessnet) were unaffected.

So I straced and gdb-ed, and I found that the crash happens in
libstdc++6.  But ldd shows the apps don't link libstdc++6.  It turned
out that the crashes happened in SCIM code, which was linked to
libstdc++6.

What happens is that gtk starts and loads SCIM modules as input methods.
Then we end up with a hybrid libstdc++5/libstdc++6 environment which
seems to cause the crashes.

I deinstalled SCIM and now the crashes disappeared.  However:
 - SCIM is a required package in various cultures using non-latin
   scripts
 - SCIM doesn't seem to be at fault, as it seems to have passed nicely
   through the gcc transition while the other crashing apps still
   haven't.

The chaos happens when GTK puts together code that doesn't know and
can't know about each other.  I don't know where to report the bug, as I
don't see any specific package being wrong here, but a generically
messed up situation.  I don't know what could be a solution either,
besides moving everything to a single kind of libstdc++.

So I write here, so that at least the mess is made known.  If someone
can see any plans for solutions, I can't :(


Ciao,

Enrico

[1] Description: Smart Common Input Method platform
--
GPG key: 1024D/797EBFAB 2000-12-05 Enrico Zini <enrico@debian.org>

Attachment: signature.asc
Description: Digital signature


Reply to: