segfault in wgenopts.c

[Rob Taylor]

Urrk. just realised I jumped in here without giving you guys any platform/ build
option info..


build is i686-pc-gnu-linux
with --enable-c-mbchar=yes --enable-cstdio=libio --enable-long-long --enable-sha
red --enable-threads.

Yes, I know, I must be mad wanting to do this...

Rob Taylor


> > I hit this. looking tah teh code it looks liek there's a load of
> problems with
> > the w* versions in libio not using the cortrect JUMPTABLE fuucntions (e.e
> > wgenopts.c uses _IO_DOALLOCATE rather than _IO_WDOALLOCATE, etc.
> > I'm makeing the changes now but I've hit a problem in wfiledoalloc.c.
> >
> > in _IO_wfile_doalloc a call is made to _IO_file_doalloc if
> > _IO_buf_base ==NULL.
> > can someone explain why this is done?
>
> Who is working on this? I commented out that piece of code, chanegd a
> subsequent
> call to _IO_SYSSTAT yo _IO_WSYSSTAT and changed the call to _IO_DOALLOCATE in
> wgenopts.c to _IO_WDO_ALLOCATE. this giot it a bit further in initialisation.
>
> I found a problem in __basic_string<wchar_t>::__basic_string - it calls
> IO_doallocbug, but it appears that the non wide entries in the
> JUMPTABLE are not
> initialised.
>
> I'm starting to get out of my depth here - commenting these calls out
> got me as
> far as finilistion (_fini) but then I hit some weirdo errors in std_fstream
> static destructors.
>
> anyone got any pointer / wants my patches?
>
> Thanks,
> Rob Taylor
> Flying Pig Systems
>