Re: (971023) Addendum: i386-linux-gnulibc1 g++ test results

[Thomas Weise <tom at zaphod dot wh9 dot tu-dresden dot de>]

Jeffrey A Law wrote:
> 
>   In message <199710240843.KAA07400@mururoa.inria.fr>you write:
>   >     May be, I can expand a little on these exception handling problems.
>   >     I'm not quite sure that the following problem is related to those
>   >     that you reported, but it might be one instance of the problem...
>   >
>   >     My system is Linux with a 2.0.29 kernel and libc.so.5.3.12.
>   >     The compiler is configured as a i586-pc-linux-gnulibc1.
>   >     The problem is consistently present with all snapshots since (and
>   >     including) 970917, I believe the example worked fine with the 1997-09-10
>   >     but I have to check it once again. My test example (excerpted from blitz++)
>   >     does not work properly.
>   >
>   > // Exceptions
>   > // BZ_EXCEPTIONS
> [ ... ]
> This works fine for me with the current development sources; you
> should try your test again when I make the next snapshot.
> 
> jeff

I've got a similiar problem, but unfortunately I'm not able to narrow it
to an appropriate test case yet. (Linux 2.0.29/libc5.4.39). The symptom
is after throwing an exception my process hangs for a time, and after
that it crashes with SIGSEGV. This may depend on the produced binary, as
I've seen it working by varying static libraries needed. (Those libs
have been compiled with exactly the same options). However, *all works
fine* rebuilding completely using -fsjlj-exceptions, what confirms that
this bug apears in snapshots later then egcs-2.90.06 when DWARF EH has
been put in.

As there should be some EH patches in next snapshot I hope they will
solve this problem as well. I would also greatly appreciate some hints
on how to provide a helpful bug report for cases like this.

-- 
Thomas Weise, http://www.inf.tu-dresden.de/~tw4
Dresden University of Technology, Department of Computer Science