Fri 28 Jan 2011 08:26:54 AM UTC, comment #11:
I was not logged in yesterday night - this issue is fixed in with the patch.
Best,
Volkmar
|
Thu 27 Jan 2011 10:28:04 PM UTC, comment #10:
It seems to work for me. Thank you very much for fixing it.
|
Thu 27 Jan 2011 10:17:05 PM UTC, comment #9:
Can you please confirm that the patch I checked in fixes the problem described in this bug report?
|
Thu 27 Jan 2011 02:48:08 PM UTC, comment #8:
please disregard comment #7, it does belong to bug #32261.
|
Thu 27 Jan 2011 01:39:07 PM UTC, comment #7:
I've uploaded the data to
Please forgive the huge amount of data and code involved. I do hope the steps to reproduce the problem are a hint as to what is going wrong.
|
Thu 27 Jan 2011 11:21:35 AM UTC, comment #6:
I checked in an additional changeset:
http://hg.savannah.gnu.org/hgweb/octave/rev/88ff30dcc048
With it, I see
So maybe this will fix the problem for you.
|
Thu 27 Jan 2011 10:03:56 AM UTC, comment #5:
I'm not sure whether the load issue was related, but it seems to be gone now. However, the original problem still persists in one way or the other. Running the attached test_job.m as a script, the field
should contain a properly initialised cfg_dep, but it doesn't. Again, the empty substructs are lost. The file test_job.mat contains the same variable, correctly saved in MATLAB and loadable in octave.
(file #22530, file #22531)
|
Thu 27 Jan 2011 09:26:40 AM UTC, comment #4:
Was that problem caused by the patch, or was it also failing to load the object before the patch was applied?
I checked in the following change.
http://hg.savannah.gnu.org/hgweb/octave/rev/a71c1aa9823e
It seems to fix the problem for me.
|
Thu 27 Jan 2011 09:02:31 AM UTC, comment #3:
There is now another problem: loading a .mat file with a cfg_dep object fails. After a cfg_dep object is created from the command line, load does not fail anymore. (see attached file)
(file #22524)
|
Thu 27 Jan 2011 01:43:40 AM UTC, comment #2:
I checked in the following change:
http://hg.savannah.gnu.org/hgweb/octave/rev/d08901c05c1b
Jaroslav, do you see a better way to solve this problem?
This seems to fix the bug for me, but I'll wait for feedback before closing the report.
|
Tue 25 Jan 2011 12:52:23 PM UTC, comment #1:
There seems to be a different behaviour in MATLAB and octave in this case, which is handled in cfg_dep/subsasgn.m. In MATLAB, cfg_dep/subsasgn is called with the following arguments:
Here, dep is not yet classified, and more importantly, it is empty.
In Octave, the situation is different.
Here, dep is not empty, but the properties do not seem to be initialised by the constructor code either. For MATLAB compatibility, dep should be empty under these circumstances.
|
Tue 25 Jan 2011 11:41:32 AM UTC, original submission:
Using the same code as in bug #32222, the following produces incorrect results:
d(1) does not get initialised properly, but d(2) does. In MATLAB, both d(1) and d(2) get initialised.
|