Ok, thanks, does this look better?

Hurm the tests seem to be failing with a latex error on windows, could it be the ylabel is too long for latex' tastes?

It seems like the test failure is unrelated. Has anyone seen that before? (Cc: @JanSchulz)

It seems latex is run in that test (if that's not ok, then the rest of my explanation is wrong :-)), but I'm not sure if appveyor has latex installed (I'm pretty sure it has not). Usually there are test decorators which test for latex and skip the test if no latex is available on the system: https://github.com/matplotlib/matplotlib/blob/master/lib/matplotlib/tests/test_backend_ps.py#L22-L25 and used then in https://github.com/matplotlib/matplotlib/blob/master/lib/matplotlib/tests/test_backend_ps.py#L72-L78

====================================================================== 
ERROR: Test for 5456. When the label was too long we'd get a ValueError that 
---------------------------------------------------------------------- 
Traceback (most recent call last):
  File "C:\conda\envs\test-environment\lib\site-packages\nose\case.py", line 197, in runTest
    self.test(*self.arg)
  File "c:\projects\matplotlib\lib\matplotlib\tests\test_tightlayout.py", line 266, in test_long_label
    plt.tight_layout()
  File "c:\projects\matplotlib\lib\matplotlib\pyplot.py", line 1282, in tight_layout
    fig.tight_layout(pad=pad, h_pad=h_pad, w_pad=w_pad, rect=rect)
  File "c:\projects\matplotlib\lib\matplotlib\figure.py", line 1857, in tight_layout
    rect=rect)
  File "c:\projects\matplotlib\lib\matplotlib\tight_layout.py", line 355, in get_tight_layout_figure
    pad=pad, h_pad=h_pad, w_pad=w_pad)
  File "c:\projects\matplotlib\lib\matplotlib\tight_layout.py", line 126, in auto_adjust_subplotpars
    tight_bbox_raw = union([ax.get_tightbbox(renderer) for ax in subplots])
  File "c:\projects\matplotlib\lib\matplotlib\axes\_base.py", line 3750, in get_tightbbox
    bb_xaxis = self.xaxis.get_tightbbox(renderer)
  File "c:\projects\matplotlib\lib\matplotlib\axis.py", line 1087, in get_tightbbox
    renderer)
  File "c:\projects\matplotlib\lib\matplotlib\axis.py", line 1070, in _get_tick_bboxes
    extent = tick.label1.get_window_extent(renderer)
  File "c:\projects\matplotlib\lib\matplotlib\text.py", line 961, in get_window_extent
    bbox, info, descent = self._get_layout(self._renderer)
  File "c:\projects\matplotlib\lib\matplotlib\text.py", line 352, in _get_layout
    ismath=False)
  File "c:\projects\matplotlib\lib\matplotlib\backends\backend_agg.py", line 230, in get_text_width_height_descent
    renderer=self)
  File "c:\projects\matplotlib\lib\matplotlib\texmanager.py", line 676, in get_text_width_height_descent
    dvifile = self.make_dvi(tex, fontsize)
  File "c:\projects\matplotlib\lib\matplotlib\texmanager.py", line 423, in make_dvi
    report))
RuntimeError: LaTeX was not able to process the following string:
'lp'
Here is the full report generated by LaTeX: 

Ah -- thanks, @JanSchulz. I don't LaTeX is supposed to be used in that test, but due to some sideeffect (probably of another test) it is.

@josecv: Can you try adding the @cleanup decorator to your new test?

Cleanup is not enough, it has to skip if no Latex is installed.

Ah, perfect, thanks! Hopefully this should do the trick

Ok -- I think the confusion is that the new test in this PR is not responsible for the failure, but it somehow triggering it in another test -- test_backend_svg:test_determinism_tex, which already has the @needs_tex decorator.

It isn't clear to me why the next test here needs to check for TeX -- it shouldn't be using it and isn't part of the traceback displayed on AppVeyor.

I've seens such behaviour when I missed a @cleanup in the tests of ggplot, but this is here: https://github.com/matplotlib/matplotlib/blob/master/lib/matplotlib/tests/test_backend_svg.py#L179-L184 -> I've no idea :-(

Huh Known Failure Did Not Fail:

====================================================================== 
ERROR: Test for 5456. No ValueError when using very long label. 
---------------------------------------------------------------------- 
Traceback (most recent call last):
  File "C:\conda\envs\test-environment\lib\site-packages\nose\case.py", line 197, in runTest
    self.test(*self.arg)
  File "c:\projects\matplotlib\lib\matplotlib\testing\decorators.py", line 152, in wrapped_callable
    func(*args, **kwargs)
  File "c:\projects\matplotlib\lib\matplotlib\testing\decorators.py", line 67, in failer
    raise KnownFailureDidNotFailTest(msg)
KnownFailureDidNotFailTest: This test needs a Tex installation  
----------------------------------------------------------------------

Is it possible that the cleanup decorator on its own is allowing the test to pass?

Well, more specifically, it did fail, but for the wrong reasons.

I am wondering if usetex=True is leaking from somewhere else?

On Wed, Mar 16, 2016 at 1:08 PM, josecv [email protected] wrote:

Huh Known Failure Did Not Fail:

ERROR: Test for 5456. No ValueError when using very long label.

Traceback (most recent call last):
File "C:\conda\envs\test-environment\lib\site-packages\nose\case.py", line 197, in runTest
self.test(_self.arg)
File "c:\projects\matplotlib\lib\matplotlib\testing\decorators.py", line 152, in wrapped_callable
func(_args, **kwargs)
File "c:\projects\matplotlib\lib\matplotlib\testing\decorators.py", line 67, in failer
raise KnownFailureDidNotFailTest(msg)

KnownFailureDidNotFailTest: This test needs a Tex installation

Is it possible that the cleanup decorator on its own is allowing the test
to pass?

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub
#6161 (comment)

Hmm, strange - hopefully then leaving the test with only @cleanup is going to do the trick

Oh my, bizarrely it's the Travis build that's failed now, though it seems to be an unrelated failure in a single test (namely TEST_ARGS=--pep8 with python 3.5):

The command "curl -s -o python-3.5.tar.bz2 https://s3.amazonaws.com/travis-python-archives/binaries/$(lsb_release -is | tr "A-Z" "a-z")/$(lsb_release -rs)/$(uname -m)/python-3.5.tar.bz2" failed and exited with 18 during .

Your build has been stopped.

And now everything passes. Go figure.

Just one thought on this. This approach solves the problem for the fully automatic case, but what about the case where the user specified some of the margins?

That's a good point - the min statements could easily be pulled out from the if not margin_*: statements, would that be the best thing to do?
The user would find that their margins are being ignored, but at least there won't be an exception. Alternatively, tight_layout could itself produce an error when fed "bad" (out of bounds) margins.

#6167 is another PR addressing the same issue. I left some comments on the other PR that I think apply here as well.

@kskod @josecv Are you both in CSC D01? If so can you please collaborate on this?

I am closing this as I believe we should instead have a more general solution which raises whenever the padding/spacing parameters make tight-layouting impossible (#8062, #8201, #8202) (or handles correctly such cases).
Feel free to ask for a reopen if you disagree.