[Boost-users] [spirit] boost 1.33.1, HP-UX ia64 & gcc-3.4.6 - doesn't work with optimization (-O2)
Frank Bergemann
some supplementary information:
The problem happens also for gcc-3.4.6 - which is the final version of gcc-3.x.x series.
If i compile for 32 bit there is even a coredump for execution:
(gdb) r
Starting program: /nfs/uh01/frank/TESTX/sample
test is '<pDeleteHlrSubscriberIn class_id="0" tracking_level="0" version="0">'
Program received signal SIGBUS, Bus error
si_code: 1 - BUS_ADRALN - Invalid address alignment. Please refer to the
following link that helps in handling unaligned data:
http://docs.hp.com/en/7730/newhelp0610/pragmas.htm#pragma-pack-ex3.
boost::spirit::grammar<basic_xml_grammar<char>,boost::spirit::parser_context<boost::spirit::nil_t>
>::grammar (this=0x40025cf0)
at sp_counted_base_gcc_ia64.hpp:38
38 "r"(pw));
I opened a bug for gcc: http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42791
And i asked HP for confirmation.
I'll keep you informed...
best regards,
Frank
___________________________________________________________
Preisknaller: WEB.DE DSL Flatrate für nur 16,99 Euro/mtl.!
http://produkte.web.de/go/02/
_______________________________________________
Boost-users mailing list
Boost...@lists.boost.org
http://lists.boost.org/mailman/listinfo.cgi/boost-users
Frank Bergemann
we have confirmation now from hp, that they could reproduce the problem
(pls. see below).
We here decided to switch to a newer gcc compiler version.
And doing so we will also switch to a newer boost lib version.
Actually we already tried the 1.41.0 with the old gcc-3.4.4.
But for that there was a not backwards compatible change in boost::spirit impl. and API.
So as we don't have time to adapt our application, i guess we need to use the 1.40.0?
cheers,
Frank
---- reply from hp ----------------------------------------------------------------------------
> I kindly would like to ask you for confirmation of the problem.
> Because i want to exclude that the problem is just due to our system
> set-up here.
> I would highly appreciate if you could try to re-produce the problem
> (see http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42791)
> and give us positive or negative confirmation.
I ran the test case from the GCC bug report using GCC 3.4.5 and Boost
1.33.1 and was able to reproduce the problem on my system.
> For decision making about new gcc compiler version to use,
> i would like to ask you for a proposal - especially related to
> performance issues.
> Because there are difference gcc version packages available from your
> download sites.
Generally, I recommend using the most recent GCC version available.
> If you confirm the problem, then it might also be helpful to add the
> information to your web-site, porting guidelines, etc.
> - for all your customers (they might hit the same problem)
> Or ask gnu to fix it.
The FSF is not going to fix anything on the 3.4.* branch, this branch is
considered 'dead' and is no longer being maintained.
[..]
[...]@cup.hp.com
______________________________________________________
GRATIS für alle WEB.DE-Nutzer: Die maxdome Movie-FLAT!
Jetzt freischalten unter http://movieflat.web.de