[rrd-developers] Seg fault with r1094

[Matthew Chambers]

I am compiling rrdtool on PPC (G4) and running it on PPC64 (G5).  This has
never been an issue before because the PPC64 very backward compatible with
PPC.  To get the debug build, I did make clean, then:
./configure --prefix=/frogstar/usr/ppc CFLAGS='-O0 -g'
Then I did make and make install.  The prefix points to a shared
arch-specific folder.  When you say send you a sample, do you mean the RRD
file that's crashing?

-Matt


> -----Original Message-----
> From: Bernhard Fischer [mailto:rep.dot.nop at gmail.com]
> Sent: Thursday, May 31, 2007 2:39 AM
> To: Matthew Chambers
> Cc: rrd-developers at lists.oetiker.ch
> Subject: Re: [rrd-developers] Seg fault with r1094
> 
> On Wed, May 30, 2007 at 07:48:20PM -0500, Matthew Chambers wrote:
> >I'm not sure if I'm doing something wrong or what.  I know when I
> submitted
> >my patch for --full-size-mode my rrdtool calls were working fine.  Since
> >then, I've discovered an unintended feature where --full-size-mode
> doesn't
> >act right with --no-legend.  In the course of trying to fix this, I
> updated
> >my working copy to the repo's version of the patch, which apparently
> >included some other changes then.  Since recompiling from then on it
> seems
> >that none of my rrdtool cards are working right.  They segfault on any
> time
> >I actually call an rrd (I've only tried info and graph, but probably any
> >subcommand).  I recompiled with -g -O0 and ran it through gdb, and I got
> >this:
> >
> >
> >
> >Starting program: /frogstar/usr/ppc/bin/rrdtool info
> >/var/lib/ganglia/rrds/__SummaryInfo__/swap_free.rrd
> >
> >
> >
> >Program received signal SIGSEGV, Segmentation fault.
> >
> >0x0ff6d1cc in rrd_close (rrd_file=0x0) at rrd_open.c:418
> >
> >418                  ret = munmap(rrd_file->file_start, rrd_file-
> >file_len);
> >
> >
> >
> >Now I'm sure I didn't change anything in rrd_open, so is this a new bug,
> or
> >am I missing something?  I just need to make sure I'm not crazy.
> 
> Sounds odd, info works fine for me.
> Can you make clean && make again, just to be sure. Also, which access
> method do you use (mmap or FD). Any additional flags you specified in
> to configure? If this still fails, i'd like to know what version of
> binutils/cc you are using, and which arch you are on.
> 
> If you are not passing anything to configure, please send a bzip2'ed
> sample to me directly.
> 
> Thanks,