From: Dan Nuffer (dnuffer_at_[hidden])
Date: 2001-02-20 13:35:26
John Maddock wrote:
> >I've never seen this before... Do you know more specific details such
> >as OS and gcc version?
> >If you have a test program I can test it out with gcc 2.95.2.
> I can reproduce this with on either BeOS or on SUSE linix (with or without
> To reproduce:
> build the library with the gcc-shared.mak make file and copy the .so output
> files to your path.
> go to the libs/regex/demo/regress directory and build:
> make -fgcc.mak debug
> run the tests:
> ./r2 tests.txt
> you should see "abort" printed out.
> If you set a breakpoint in reg_expression<char>::fail, you should see that
> this is called several times - some from within the exe, and once from
> within the .so (the ultimate caller is RegEx::RegEx I think), it's the
> latter occurance that fails. I've been meaning to try moving the throw
> statement into some non-template code in the library, and see if that fixes
> it, but I haven't had a chance yet...
> - John Maddock
Thanks for your quick reply. Unfortunately, I was able to reproduce
this. So I guess the regex++ shared library is out for now.
FYI, I also saw segfaults when I used any of the following optimization
flags: -fforce-mem -fomit-frame-pointer -fdefer-pop -O -O2 -O3 -m486.
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk