multithreaded EH is still broken on Linux/Alpha and other platforms

[Sai-Lai Lo]

>>>>> Alexandre Oliva writes:

> Sai-Lai Lo <S.Lo@orl.co.uk> writes:
>> Could you look into the problem with multithreaded EH before egcs 1.1 is
>> officially released?

> Just to make sure... Did you --enable-threads when you configured
> egcs?

Thank you very much for the hint. Indeed I have not --enable-threads.
Rebuild egcs with the flag and the problem goes away.

I'm now able to build and run omniORB2 (a multithread C++ ORB) and its
testsuite without optimization.

However, when I specify -O2, there are loads of warnings about variables
might be used uninitialized. These warnings do not appear with -O2 on x86
linux. The warnings include 'struct cp_eh_info * __exception_info' which
looks like it is EH related.

All the resulting binaries core dump in a global constructor. I'll look
into what is wrong in the global constructor some time. Does this ring a
bell?

Any suggestion as to what to look into is much appreciated.

This problem occurs both with --enable-haifa and --disable-haifa.

The test was done with egcs_980802.

Regards,

Sai-Lai Lo


-- 
Sai-Lai Lo                              |       Research Scientist
                                        |
E-mail:         S.Lo@orl.co.uk          |       Olivetti & Oracle Research Lab
                                        |       24a Trumpington Street
Tel:            +44 223 343000          |       Cambridge CB2 1QA
Fax:            +44 223 313542          |       ENGLAND