Closed GoogleCodeExporter closed 9 years ago
This indicates that an HDF5 handle is still open from the previous write.
Original comment by miller.lucas
on 9 Nov 2010 at 6:31
Running this through a debugger, it looks like all the handles are closed when
the ArchiveWriter is going out of scope.
However, there is a mysterious call to H5Fflush before H5Fclose. This extra
call should be gratuitous, since there are no open handles to flush, but
removing it causes problems in the test suite.
This may be indicative of a deeper problem.
Original comment by miller.lucas
on 9 Nov 2010 at 9:18
Hmm. I am unable to reproduce the following behavior, as reported by Mark
Davies:
"I can get a valid hdf5 file. The steps are write out the first file it
will be bad, attempt to overwrite the .abc file it will fail, then
write the same data to a new file and it will be ok."
or:
"However, if I give it a frame range it it will say that everything is
fine, i.e. tell me the frames that are written and that it closed the
archive, but the resulting file is not valid. The hdf5 file cannot be
opened by the tools h5stat, h5dump, h5debug etc, they all fail on
H5F_open()."
I've tried this with exporting the full frame range or not the full frame
range. My details are:
Maya 2010
HDF5 1.8.5-patch1
Boost 1.44
The scene is a polymesh sphere with animated transforms (translations).
Would it be possible to send me a test Maya scene?
Original comment by ard...@gmail.com
on 9 Nov 2010 at 11:31
I sent Joe an example of a deforming poly sphere which recreates Mark's problem.
I have a potential solution that I think fixes both of your issues.
For more details see:
http://code.google.com/r/millerlucas-dev/source/detail?r=3d72463d0e36f5841c272f6
31d30ae0af6917dff
Original comment by miller.lucas
on 10 Nov 2010 at 2:05
Original comment by scottmmo...@gmail.com
on 20 Nov 2010 at 2:35
This is fixed in the clones ardent-embic and millerlucas-dev. It will be
pushed to the mainline in the next couple days (both the hg repo and the source
download).
Original comment by ard...@gmail.com
on 6 Dec 2010 at 11:53
Original issue reported on code.google.com by
sven...@gmail.com
on 9 Nov 2010 at 6:22Attachments: