Use v3.1 imagehash.hex_to_hash #2240
Conversation
|
We've got some looooong running tests, that are starting to kill us ... something reeks within. |
| @@ -776,7 +761,9 @@ def _hex_to_hash(hexstr, hash_size=_HASH_SIZE): | |||
| else: | |||
| uris = repo[unique_id] | |||
| # Create the expected perceptual image hashes from the uris. | |||
| expected = [_hex_to_hash(os.path.splitext(os.path.basename(uri))[0]) | |||
| to_hash = imagehash.hex_to_hash | |||
There was a problem hiding this comment.
Bit odd... does this keep the next line <80char in length?
| def _calculate_hit(uris, phash, action): | ||
| # Create the expected perceptual image hashes from the uris. | ||
| expected = [_hex_to_hash(os.path.splitext(os.path.basename(uri))[0]) | ||
| to_hash = imagehash.hex_to_hash |
| @@ -107,7 +107,7 @@ install: | |||
|
|
|||
| script: | |||
| - if [[ $TEST_TARGET == 'default' ]]; then | |||
| python -m iris.tests.runner --default-tests --system-tests --print-failed-images; | |||
| travis_wait python -m iris.tests.runner --default-tests --system-tests --print-failed-images; | |||
There was a problem hiding this comment.
Does this change belong here..?
There was a problem hiding this comment.
Just testing this out ...
bjlittle
force-pushed
the
update-imagehash
branch
from
376e9a9 to
eaece3f
Compare
|
Should just strip those last two... |
bjlittle
force-pushed
the
update-imagehash
branch
from
f137282 to
f09db41
Compare
bjlittle
force-pushed
the
update-imagehash
branch
from
f09db41 to
bd9a368
Compare
| @@ -163,7 +163,7 @@ def run(self): | |||
|
|
|||
| args = ['', None, '--processes=%s' % n_processors, | |||
| '--verbosity=2', regexp_pat, | |||
| '--process-timeout=600'] | |||
| '--process-timeout=180'] | |||
There was a problem hiding this comment.
@dkillick This change brought the TEST_TARGET=default matrix runners from +18m down to 6m17s for py27, and 7m48s for py34, whoop! ... that's at least 10 minutes quicker than I've seen for a while ... 🎉
Admittedly, it was me that bumped it up that high due to the recent graphics testing changes 😱 ... looks like nose uses the full timeout per multi-process runner regardlessly. Originally I thought this was used as a timeout hwm ... clearly not?! If you watch the travis tests running live, you'll notice that the last test finishes, but there is a period of delay before the test runner terminates. The process timeout appears to be used by nose as a process sample time also.
Clearly there is a sweet-spot for this process-timeout for our tests, which we may have to tweak, but I'm totally 👍 for this change!
This change also resolves the following carnage that other PRs are experiencing in travis at the moment:
No output has been received in the last 10m0s, this potentially indicates a stalled build or something wrong with the build itself.
Check the details on how to adjust your build configuration on: https://docs.travis-ci.com/user/common-build-problems/#Build-times-out-because-no-output-was-received
10m0s or 600s is the magic number that's causing this problem ... what were the odds!
There was a problem hiding this comment.
This number sounds suspiciously like the default image tolerance from the image tests... Still, we can change the value as we need, and always put some effort into fixing it in a year or two 😉
And it's great to get the travis tests running faster! Good stuff @bjlittle 👍
Now that SciTools/conda-recipes-scitools#210 has been merged (thanks to @ocefpaf in SciTools/conda-recipes-scitools#211) we can now use the shiny new
v3.1ofimagehashwhich has the fix required toimagehash.hex_to_hashfor variablehash_size's.