[Simgrid-user] Re: compile CVS problem

quinson at loria.fr quinson at loria.fr
Wed Jul 18 02:59:48 CEST 2007


Hello Soumia,

I'm sorry I didn't the mail you sent me privately, I have some mail issues
forcing me to use a webmail...

The problem you report in the CVS (along with a ton of similar ones) were fixed
today. A cvs up should suffice. But there is so much going on in the CVS that
if the problem persist, do not hesitate to run "make clean" from the root, or
even to scratch your tree and checkout a fresh one.

For the record, this problem is a false positive. Indeed, the variable can be
uninitialized (since it's initialized in a TRY/CATCH block), but if the
initialization function raises an exception, we rethrow it after adding some
more info. So, there is nothing bad in this chunk of code, after all.

The interesting part is that gcc understands the trick when compiling with -O0,
but gets fooled somehow at -03...

This, by the way, explains that I didn't saw the issue before commiting this
"broken" code. I always compile my tree with --disable-compile-optimization so
that the assembly code match closer the source code, making my life with gdb
and valgrind more pleasant. Adding --with-pthreads may also help to debug
sometimes.

And lastly, this kind of discussion may be more appropriate on -devel, to which
you definitively should subscribe.

Bye, Mt.




More information about the Simgrid-user mailing list