Ulrich,

I'll recompile gcc and let you know.  While trying to get Qt to compile (and
not segfault), I deleted the working directories to make space on my disks.

If you haven't heard me comment on this before, I'll do it now.  For the
last two versions of gcc (3.2 and 3.3), I've gotten a _lot_ of cases where
code will compile and then segfault that used to run fine when compiled with
previous versions of gcc.  I'm especially seeing problems in applications
that use threading.  Sometimes, even specifying -O0 doesn't help.  (And in
the case of glibc, you _can't_ specify -O0.  The package complains that you
can't compile it without optimization and quits.)


Mark Post

-----Original Message-----
From: Ulrich Weigand [mailto:[EMAIL PROTECTED]
Sent: Thursday, July 24, 2003 4:56 PM
To: [EMAIL PROTECTED]
Cc: [EMAIL PROTECTED]
Subject: Re: Problem(s) with gcc 3.3


Mark Post wrote:

>I compiled gcc 3.3 on my LCDS instance.  When I ran the "make check"
command
>it failed on the gctest.  This seems to be a test of the Boehm garbage
>collector.

Well, it works for me ;-)  Can you try re-running gctest manually; go to
the s390-*-linux/boehm-gc/ subdirectory of the build tree and run ./gctest.
I get:

[EMAIL PROTECTED] boehm-gc]$ ./gctest
Completed 3 tests
Allocated 5586007 collectable objects
Allocated 306 uncollectable objects
Allocated 3738956 atomic objects
Allocated 32958 stubborn objects
Finalized 6586/6586 objects - finalization is probably ok
Total number of bytes allocated is 183059492
Final heap size is 15953920 bytes
Collector appears to work
Completed 104 collections

>This makes me somewhat concerned about the rest of the compiler, and what
>other problems may be lurking.  I'm hoping Ulrich and his co-workers can
>help me figure out what might be wrong.

I wouldn't necessarily consider boehm-gc to be representative of the
compiler; this package needs to do a number of 'interesting' things
in order to make garbage collection work for a language that was never
designed for it.  boehm-gc makes various assumptions about its
environment (kernel, libc, ...) and tends to break if those are
violated.  (In addition, boehm-gc is used only by the Java runtime
library, so if you don't use Java, you needn't care anyway ...)

If you complete the test suite run (make -k check) and generate a test
report (<srcdir>/contrib/test_summary), do the results look similar to
what you should expect from a 3.3 build?  E.g. compare:
http://gcc.gnu.org/ml/gcc-testresults/2003-07/msg01260.html

Bye,
Ulrich

--
  Dr. Ulrich Weigand
  [EMAIL PROTECTED]

Reply via email to