[Gc] [w3m-dev-en 00993] Re: Solaris 8 build of 0.4.2 (fwd)
hans.boehm at hp.com
Wed Apr 7 15:22:21 PDT 2004
Can you try that again with a clean build: Unpack the tar file
into an empty directory and start again. GC_debug_header_size
is new in 6.3alpha5, but I think it's pretty clearly defined in
dbg_mlc.c. If that doesn't work, can you send the build log?
Linux is my standard development environment, so things usually work there.'
> -----Original Message-----
> From: Steven F. Killen [mailto:skille1 at umbc.edu]
> Sent: Wednesday, April 07, 2004 2:35 PM
> To: Boehm, Hans
> Cc: gc at napali.hpl.hp.com
> Subject: RE: [Gc] [w3m-dev-en 00993] Re: Solaris 8 build of
> 0.4.2 (fwd)
> On Wed, 7 Apr 2004, Steven F. Killen wrote:
> > > I think we really need the stack trace at the segmentation fault.
> > > It's possible the GC_LOOP_ON_ABORT environment variable might help
> > > if there's otherwise a problem extracting that from the debugger.
> > > (Wait until you're sure its looping in the SIGSEGV
> handler, the attach
> > > a debugger.)
> > As a sanity check I pulled down the recent release alpha5
> after I noticed
> > it, and it seems to pass make check just fine. I didn't
> see anything
> > immediately related to my problem in the changelog, but if
> you still want
> > a stack trace I'll see about that. Go figure. :)
> I should not have celebrated so quickly. IRIX and Solaris
> pass make check
> on alpha5, but now linux has issues:
> gmake check-TESTS
> gmake: Entering directory `/usr/src/skille1/gc/000/gc6.3alpha5'
> lt-gctest: relocation error: lt-gctest: undefined symbol:
> FAIL: gctest
> Could you give a quickie summary on how to do a stack trace?
> I'm not too
> familiar with it, but would like to help.
> Steve Killen <skille1 at umbc.edu>
More information about the Gc