Wed 23 Aug 2017 08:39:51 AM UTC, comment #15:
Thanks, I pushed the patch after removing trailing whitespaces.
Closing report as fixed.
|
Tue 22 Aug 2017 08:07:46 PM UTC, comment #14:
@Pantxo: Seems like the patch is ready to commit. Can you strip trailing whitespace before you do so? I noticed an extra space at the end of a line in the docstring for subplot.
|
Tue 22 Aug 2017 08:59:12 AM UTC, comment #13:
@Pantxo: see comment #11 (checked with the old patch, after adding the missing x, on Win-10 machine)
|
Tue 22 Aug 2017 08:32:36 AM UTC, comment #12:
@Avinoam: indeed I forgot to add the "x" after copy/paste. I attached a modified patch.
(file #41614)
|
Tue 22 Aug 2017 07:42:37 AM UTC, comment #11:
The results looks good, see attached file.
Checking other demos, I did not see problem
(file #41610)
|
Mon 21 Aug 2017 10:28:13 PM UTC, comment #10:
There seems to be a typo in the patch
should be
(deftypefnx and not deftypefn)
|
Mon 21 Aug 2017 09:34:57 AM UTC, comment #9:
The problem is that "newplot" now removes all listeners from the axes, including those needed for subplot alignment.
I attached a patch that fixes this issue for me:
- subplot.m is modified to include a "subplot (R,C, IDX, HAX)" call form: this is ML compatible; it turns HAX into a subplot.
- newplot calls subplot to reinstall listeners after resetting the axes.
(file #41605)
|
Thu 10 Aug 2017 05:40:33 PM UTC, comment #8:
With Doug's fingering of the problem cset I went and compiled revision 23012 and 23013, and I also can verify that 23013 introduced the problem.
|
Thu 10 Aug 2017 05:15:36 PM UTC, comment #7:
yes ccache
I start going back by powers of 2 ie.
back 128 and if still bad go
back 256 etc.
powers of 2 is most efficient, as you know.
17 min for a full ./bootstrap to ./run-octave.
most of the time much less.
|
Thu 10 Aug 2017 05:09:10 PM UTC, comment #6:
Are you using ccache as well?
Your computer, indeed, must be very fast. A 7-month old bisection would typically take me close to a day to perform.
|
Thu 10 Aug 2017 05:05:13 PM UTC, comment #5:
I just copied #7 and made an mfile.
I did this manually I did not use a bisect command if there is one.
:-)
I also have a fast computer
|
Thu 10 Aug 2017 04:45:43 PM UTC, comment #4:
I am curious -- what did you use as test for bisect?
Dmitri.
--
|
Thu 10 Aug 2017 04:07:34 PM UTC, comment #3:
@Doug: Wow, thanks for doing the bisection. That is a long time a go indeed. I'm adding Pantxo to the CC list for this bug since he checked in cset 23013. I happen to know that he is on summer holiday right now, so don't expect an answer for a few weeks.
|
Thu 10 Aug 2017 03:58:27 PM UTC, comment #2:
I bisected to
Changeset:
23013 (5dca1727ce94) cla.m, newplot.m: make use of reset function to cleanup listeners (bug #49980). …
this 7 months go!
|
Thu 10 Aug 2017 04:52:23 AM UTC, comment #1:
Confirmed. I don't run the demos all that frequently, but this feels like a fairly new regression. Maybe try hg bisect to find the problem changeset?
|
Thu 10 Aug 2017 04:02:50 AM UTC, original submission:
margin between plot and colorbar is incorrect (in some orientations).
See e.g. "demo colorbar". Demo #7 is attached. Good plot made by octave-4.2.1, bad on by hg id cb8251337689.
Dmitri.
--
|