[rrd-developers] MRTG , Segmentation Fault , After re-installing perl .

[Tobias Oetiker]

the part of the code giving trouble is pretty sure to cause a
segfault for all calls to rrd_update on platforms where time_t is
NOT 64bit ... which is still pretty common I guess ...

I will publish 1.3.7 as soon as I get some more feedback that the
patch actually fixes the problem others have seen.

cheers
tobi


Today kevin brintnall wrote:

> On Wed, Apr 01, 2009 at 11:01:24AM -0800, Mr. James W. Laferriere wrote:
> >    	Hello Kevin ,
>
> Hi James, forgive my late reply..  I am out on holiday.
>
> >  	There are two places in the make process that that file is compiled or
> > built ,  They're in the attached file , make-rrdtool-1_3_6.rrd_update_c.log .
> >  	I also have a full make log capture if you ever need that .
>
> These do look good..  -g and no -O.  However, based on the backtrace it
> does look like the call to write_RRA_row() is still being optimized out.
>
> Does your build environment contain any other things that would turn on
> optimization?  i.e. $CFLAGS in the environment or in the default make
> settings?
>
> Send me the full log off-list.
>
> > > It will be very helpful to see the ds_idx value; I suspect it's this
> > > pointer de-reference that's causing the SIGSEGV, because all the other
> > > pointers look correct: 	rrd->ds_def[ds_idx].ds_nam
> >
> >  	See attached ,  rrdtool-segfault-20090401.log .
>
> Your debug commands:
>
> 	(gdb) p rrd->stat_head->ds_cnt
> 	No symbol "rrd" in current context.
> 	(gdb) p rrd->stat_head
> 	No symbol "rrd" in current context.
> 	(gdb) p rrd->ds_def[ds_idx].ds_nam
> 	No symbol "rrd" in current context.
> 	(gdb) exit
> 	Undefined command: "exit".  Try "help".
>
> ...are not useful since they are in the wrong stack frame.  You need to
> use "(gdb) select 4" to examine this frame.  It would be helpful if you
> could retry.
>
> 	#4  0xb7cc1193 in write_to_rras (rrd=0xbff20944, rrd_file=0x897b658, rra_step_cnt=<value optimized out>, rra_begin=2760,
> 	    current_time=1238611056, skip_update=0x89cdfb0, pcdp_summary=0xbff209a0) at rrd_update.c:1944
>
> Also, see Tobi's recent patch.  This is the same part of code that is
> giving you trouble.
>
> 	https://lists.oetiker.ch/pipermail/rrd-developers/2009-April/003074.html
>
> What are sizeof(time_t) and sizeof(long long) on your platform?
>
>

-- 
Tobi Oetiker, OETIKER+PARTNER AG, Aarweg 15 CH-4600 Olten, Switzerland
http://it.oetiker.ch tobi at oetiker.ch ++41 62 775 9902 / sb: -9900