Re: [ooc-compiler] Segmentation fault during installation

[Tim T. <ra...@ed...>]

Hallo!

Another suggestion. What threading library do you use. Suse 9.0 might be =

prepared for Kernel 2.6.x and thus might not have choosen the classic=20
pthreads package but the new three letter acronym version (LWT or=20
something like that). This may interfere witht he gc. You should be able =

to find that out by doing an ldd on the libgc.so.

You should also do an ldd on the oo2c binary that is used during the=20
build to find out if it is linked against the gc, but I think you=20
already have proven that.

Is the crash reproducable? Do you always get the same call stack when=20
you try to compile (using gdb) or is the call stack random. If it is=20
random, Michaels guess that this is a problem with the gc is more=20
likely, if is always stays the same, it is more likely that there is a=20
bug in the compiler (speaking of 60/40 or 70/30 and reverse). Ahhh,=20
Michael, did you use that special gc method to make the gc initialize=20
memory with a specific pattern to check for non-initialized variables?

How do you compile install the gc? Do you use one of the special=20
makefiles or the configure script? If you use the configure script=20
perhaps send the output (packed) to the mailinglist. Stewart, does it=20
generate a config.h, where we could detect strange configurations?

As Stewart hinter, try to explicitely compile the gc (and oo2c) single=20
threaeded, if multi-threaded is the default.


I do not have Suse 9.0 in reach. I have compiled the oo2c on Sun=20
Solaris, Windows, Mac OS X, and various Debian versions (woody and=20
testing) on x86 and ppc processors and never had a problem. Strange...

--=20
Gru=DF...
        Tim.