Re: GtkWindow fails to pass make test

["muppet" <scott asofyet org>]

Jeff Trefftzs said:
> 3.  Gtk2-0.94 is breaking.  The process to date has been:
>       perl Makefile.PL PREFIX=~/myperl   [Succeeds with no errors]
>       make   [Works, mostly.  I'm on a RH9 system, and I've noticed that, on
> large compiles, gcc segfaults somewhere in the middle, but picks up
> where it left off in I simply say 'make' again, and things work.  This
> time it broke while compiling GtkWindow.c, though.  So I deleted
> xs/GtkWindow.c, xs/GtkWindow.o and said 'make' again, and it seemed to
> be happy.]

are you running OpenMosix?  ross ran across this sort of behavior for gcc when
running an OpenMosix kernel, and the fix was "don't do that".

i've also seen problems with gcc segfaulting while compiling a kernel on my
wife's inspiron laptop --- because the kernel that shipped with her distro
didn't include support for the machine's fan.  the annoying thing is that the
compile would cause the machine to overheat and miscompile things until it
just locked up.  (the hi-tech fix was to compile with the laptop suspended
over the A/C vent.)

in other words, if you're having GCC segfaults and seeing test errors that say
they shouldn't happen, i can't really trust that you're getting an
accurately-compiled Gtk2.so.


before release i tested the build on
   RH 8.0, gtk+-2.0.6
   RH 8.0, gtk+-2.2.1
   Win2k, dropline's gtk+ 2.2 binaries

i just re-ran the test on RH8.0 / gtk+ 2.2.1 and i can't replicate your problem.


> Solidly repeatable.  All the remaining tests were successful.

i am guessing the perl tests are dying in the same place because the compiled
code is broken in the same spot.  if you recompile and relink everything, do
you get the same results?  (i know it takes forever to compile, i
apologize...)

-- 
muppet <scott at asofyet dot org>