tests/eas fixes #526
tests/eas fixes #526
Conversation
tests/eas/generic.py
Outdated
| @@ -29,8 +29,6 @@ | |||
| from unittest import SkipTest | |||
|
|
|||
| import matplotlib.pyplot as plt | |||
| import numpy as np | |||
| import pylab as pl | |||
| import os | |||
There was a problem hiding this comment.
Since you are at it you could do a general cleanup of the imports, which means having something like:
import one
import two
# etc.
from foo import bar
# etc.
valschneider
force-pushed
the
tests/fix-plot-placement
branch
from
11dd7bf to
cefd9ef
Compare
libs/utils/env.py
Outdated
| @@ -525,6 +525,7 @@ def _init_target(self, force = False): | |||
| self.target = devlib.AndroidTarget( | |||
| platform = platform, | |||
| connection_settings = self.__connection_settings, | |||
| working_directory = '/data/local/tmp/devlib-target', | |||
There was a problem hiding this comment.
While looking at this, I remembered we had already a 'workdir' configuration option, which is configurable via target.conf for example, which AFAIR was there exactly for that scope... however, perhaps due to some refactoring, it's now defined but never used 👎
You can see it's definition at line 224 above.
Can you please wire this up to this existing configuration option?
Mind that there is also a WORKING_DIR_DEFAULT, which I don't like so much.... we should probably instead have something like:
elif platform_type.lower() == 'android':
# Workaround for ARM-software/devlib#225
self.workdir = conf.get('workdir', '/data/local/tmp/devlib-target')
self._log.debug('Setup ANDROID target...')
self.target = devlib.AndroidTarget(...)
#...
else:
raise ValueError('Config error: not supported [platform] type {}'\
.format(platform_type))
self.workdir = self.target.working_directorywhere this last statement is just to ensure to always have the two variables aligned.
There was a problem hiding this comment.
In the example above... of course in the Android case you pass in self.workdir as working_directory in the AndroidTarget ctor ;-)
There was a problem hiding this comment.
Devlib already defines default working directories for each of its target types - could I just drop the WORKING_DIR_DEFAULT thing ? That way I can just make self.workdir default to None and let devlib handle it.
tests/eas/generic.py
Outdated
| @@ -322,7 +319,7 @@ def plot_expected_util(self, experiment, util_df): | |||
| prev = time | |||
|
|
|||
| figname = os.path.join(experiment.out_dir, 'expected_placement.png') | |||
| pl.savefig(figname, bbox_inches='tight') | |||
| fig.savefig(figname, bbox_inches='tight') | |||
There was a problem hiding this comment.
This should better be on a different patch... either by itself, or eventually squased into "tests/eas/generic: Close exp_placement figure" which seems another fix... in this last case please update the commit message too.
There was a problem hiding this comment.
AFAIR this change allows us to drop the pylab import, hence why it's in this commit.
There was a problem hiding this comment.
Well, in that case, since we already use pl.savefig all around in the analysis modules... perhaps removing the pylab module perhaps should be a self contained patch which does the same all around. Do you think that's possible? At least we keep consistency on how we save generated plots...
There was a problem hiding this comment.
Ok, I'll keep that for another PR.
Removes a duplicate import and re-arranges the order of the imports.
It seems like closing unused figures isn't a default behaviour, and we are getting some warning about that.
tests/eas/ was missing an __init__.py, which meant the backend set in
tests/__init__.py wouldn't be used.
This commit also removes the fill_between() call that is used to draw
the expected_placement figure, as it seems the desired behaviour is
not supported with the Agg backend ("Unknown property step").
Seems like the use of `self.workdir` got thrown away at some point. This commit reconnects it to the devlib target constructor, and also gets rid of the `WORKING_DIR_DEFAULT` definition since devlib already defines default workdirs for its targets.
As mentionned in ARM-software/devlib#225, using /sdcard causes some issues with the HiKey960. As we haven't had any issues with using /data/local/tmp/ as a devlib working directory, we now the use this directory by default.
valschneider
force-pushed
the
tests/fix-plot-placement
branch
from
cefd9ef to
c21eacc
Compare
|
| @@ -306,7 +306,6 @@ def plot_expected_util(self, experiment, util_df): | |||
| color = 'red' | |||
|
|
|||
| tdf.plot(ax=ax[cpu], drawstyle='steps-post', title="CPU{}".format(cpu), color=color) | |||
| ax[cpu].fill_between(tdf.index, tdf, 0, step='post', color=color) | |||
There was a problem hiding this comment.
Does it make sense to test and use the method if the backend supports it?
I guess this allows to get better pretty locking plots when we run tests from out desktop....
There was a problem hiding this comment.
I had a quick look for that and it didn't seem very intuitive - that "step" parameter is part of the fill_between API so imho it should be supported by all of them. I have a post-it somewhere to remind me to look at that again sometime in the future...
There was a problem hiding this comment.
What about just a simple try: except pass?
There was a problem hiding this comment.
Not too hot on that but we could do it.
I guess this allows to get better pretty locking plots when we run tests from out desktop....
For that you'll also have to change the backend defined in tests/__init__.py
There was a problem hiding this comment.
As per chat we had, seems that fixing this is not enough to allow the removal of the Agg backend.
So, let's got with this version.
This PR brings some fixes to the expected_placement plotter in generic.py and also restores the old working directory for AndroidTarget as a temp fix to ARM-software/devlib#225