On Thu, 2 Feb 2006, Mark Taylor wrote:
> C [libc.so.6+0x725a9]
> C [libc.so.6+0x71d59]
> C [libc.so.6+0x710fd] malloc+0x8d
> C [libjnihds.so+0x31f07] rec_alloc_mem+0x3f
> C [libjnihds.so+0x28357] dau_refill_flq+0x23
do we know which line?
> C [libjnihds.so+0x1518d] dat1_alloc_lcp+0x151
> C [libjnihds.so+0x215c2] datCell+0xea
> C [libjnihds.so+0x9a95] Java_uk_ac_starlink_hds_HDSObject_datCell+0xcd
> j uk.ac.starlink.hds.HDSObject.datCell([J)Luk/ac/starlink/hds/HDSObject;+0
> j uk.ac.starlink.hds.ARYReadChannel.source()Ljava/lang/String;+48
>
> Tim, I see that you've made some changes to HDS in the last 24 hours -
> ring any bells?
>
Would this have run with my 11pm commits? Or yesterday's version?
I've patched dau_refill_flq to track all the mallocs so that when hdsStop
is called all the memory is freed. This is important for valgrind so that
I can focus on real malloc issues rather than unimportant ones (since in
most cases hdsStop is only called once at program exit).
I'll wait and see whether it fails again tonight. (it all seems to work in
my tests without any valgrind errors).
--
Tim Jenness
JAC software
http://www.jach.hawaii.edu/~timj
|