Ok, I've compiled 1.8.5 with CFLAGS=-g and ran my battery of Silo tests
on it. I am hitting three assertion failures. Two of these I have
reported in the past on 1.8.2 and 1.8.4. The third one is a new one. The
are described below. Part of the difficulty in providing information is
that I think these occur in the I/O pipeline and so the stack trace is a
bit meaningless to me. Any hints on what I could do to help get better
stack traces?
Also, be aware that I use HDF5-1.8 series libs using the
#define H5_USE_16_API mode of compilation.
Assertion failure #1...
This is occuring in a test where I am using my own compression filters.
My 'can_apply' method has been set to always return 1 as I have had
difficulty if 'can_apply' returns '0'. So, I rely solely on the op
method returning '0' if it cannot compress and do not rely on the
can_apply method. In the specific test that triggers this assertion, my
compression filter attempts to compress by 1000:1 ratio and fails
because it cannot achieve it. So, it returns a value of 0 and DOES NOT
TOUCH buf_size of buf arguments of the op caller.
#0 0x00163cef in raise () from /lib/tls/libc.so.6
#1 0x001654f5 in abort () from /lib/tls/libc.so.6
#2 0x0015d619 in __assert_fail () from /lib/tls/libc.so.6
#3 0x00852d93 in H5D_close (dataset=0x9eb2118) at H5Dint.c:1419
#4 0x00c13ee1 in FUNC.11 () from
/home/miller86/tmp/hdf5-1.8.5-pre2/install/lib/libhdf5.so.6
#5 0x00c13bd7 in FUNC.11 () from
/home/miller86/tmp/hdf5-1.8.5-pre2/install/lib/libhdf5.so.6
#6 0x00000551 in ?? ()
#7 0x00c13eac in FUNC.11 () from
/home/miller86/tmp/hdf5-1.8.5-pre2/install/lib/libhdf5.so.6
#8 0x09eb2d28 in ?? ()
#9 0x00000000 in ?? ()
Assertion failure #2...
This is failing in an H5Aread call. That same call, with identical
arguments, succeeds a few times prior to this failure.
#0 0x00138cef in raise () from /lib/tls/libc.so.6
#1 0x0013a4f5 in abort () from /lib/tls/libc.so.6
#2 0x00132619 in __assert_fail () from /lib/tls/libc.so.6
#3 0x00de6e4f in H5T_conv_s_s (src_id=27984512, dst_id=156518872,
cdata=0x9544d4f, nelmts=14515311, buf_stride=50332219,
bkg_stride=50332220, buf=0x9544a34, bkg=0x1, dxpl_id=0) at H5Tconv.c:4451
#4 0x019382e4 in FUNC.339 () from
/home/miller86/tmp/hdf5-1.8.5-pre2/install/lib/libhdf5.so.6
#5 0x0192c233 in FUNC.17 () from
/home/miller86/tmp/hdf5-1.8.5-pre2/install/lib/libhdf5.so.6
#6 0x00001163 in ?? ()
#7 0x0192c4f7 in FUNC.17 () from
/home/miller86/tmp/hdf5-1.8.5-pre2/install/lib/libhdf5.so.6
#8 0x094ed848 in ?? ()
#9 0xa040f5a8 in ?? ()
#10 0x09518b68 in ?? ()
#11 0x09540ad8 in ?? ()
#12 0x00000000 in ?? ()
This is a new failure that does NOT occur in 1.8.4. It is happening in a
test where many files are being opened and closed. I have yet to
investigate further to find out where it is precisely failing.
#0 0x088d3cef in raise () from /lib/tls/libc.so.6
#1 0x088d54f5 in abort () from /lib/tls/libc.so.6
#2 0x088cd619 in __assert_fail () from /lib/tls/libc.so.6
#3 0x00356e1d in H5C_load_entry (f=0x5827be4, dxpl_id=0, type=0xb22ab008,
addr=29234605433298983, udata=0xa000008,
skip_file_checks=18464604) at H5C.c:8250
#4 0x0068fdd4 in FUNC.88 () from
/home/miller86/tmp/hdf5-1.8.5-pre2/install/lib/libhdf5.so.6
#5 0x0067cef9 in FUNC.13 () from
/home/miller86/tmp/hdf5-1.8.5-pre2/install/lib/libhdf5.so.6
#6 0x00002027 in ?? ()
#7 0x0067dcb8 in FUNC.13 () from
/home/miller86/tmp/hdf5-1.8.5-pre2/install/lib/libhdf5.so.6
#8 0xfffffcdb in ?? ()
#9 0xffffffff in ?? ()
#10 0x06e27be4 in ?? ()
#11 0x00000000 in ?? ()
On Fri, 2010-05-28 at 13:54, Mike McGreevy wrote:
> Hi all,
>
> A pre-release version of HDF5 1.8.5 is available for testing, and can be
> downloaded at the following link:
>
> ftp://*ftp.hdfgroup.uiuc.edu/pub/outgoing/hdf5/hdf5-1.8.5-pre2/
>
> If you have a few moments to test this, we would greatly appreciate it.
> While we have tested it on the platforms that we have access to, there
> are many others that we are unable to test on ourselves. Please report
> any errors to [email protected], preferably within the next week or so.
> We are planning to create and distribute the official release by
> mid-June if no critical errors are reported beforehand.
>
> Thank you!
>
> The HDF Team
>
> _______________________________________________
> Hdf-forum is for HDF software users discussion.
> [email protected]
> http://*mail.hdfgroup.org/mailman/listinfo/hdf-forum_hdfgroup.org
--
Mark C. Miller, Lawrence Livermore National Laboratory
================!!LLNL BUSINESS ONLY!!================
[email protected] urgent: [email protected]
T:8-6 (925)-423-5901 M/W/Th:7-12,2-7 (530)-753-851
_______________________________________________
Hdf-forum is for HDF software users discussion.
[email protected]
http://mail.hdfgroup.org/mailman/listinfo/hdf-forum_hdfgroup.org
