194707 2019-02-15 09:19:00 -0800 pytest is not correctly auto-installed 2019-02-25 12:54:05 -0800 1 1 1 Unclassified WebKit Tools / Tests WebKit Local Build Unspecified Linux RESOLVED FIXED InRadar P2 Normal --- 1 psaavedra psaavedra ap bburg cgarcia commit-queue ews-watchlist glenn jbedard lforschler mcatanzaro webkit-bug-importer oldest_to_newest 1506590 0 psaavedra 2019-02-15 09:19:00 -0800 ``` python ./Tools/Scripts/run-webdriver-tests --json-output=webdriver_tests.json --release --gtk in dir /home/buildbot/webkitgtk/gtk-linux-stable4-64/build (timeout 1200 secs) Using port gtk Test configuration: <, x86, release> Using display server xvfb Using driver at /home/buildbot/webkitgtk/gtk-linux-stable4-64/build/WebKitBuild/Release/bin/WebKitWebDriver Browser: MiniBrowser Parsing expectations Traceback (most recent call last): File "./Tools/Scripts/run-webdriver-tests", line 86, in <module> runner.run(args) File "/home/buildbot/webkitgtk/gtk-linux-stable4-64/build/Tools/Scripts/webkitpy/webdriver_tests/webdriver_test_runner.py", line 70, in run runner_tests = [runner.collect_tests(tests) for runner in self._runners] File "/home/buildbot/webkitgtk/gtk-linux-stable4-64/build/Tools/Scripts/webkitpy/webdriver_tests/webdriver_test_runner_selenium.py", line 51, in collect_tests executor = WebDriverSeleniumExecutor(self._driver, self._env) File "/home/buildbot/webkitgtk/gtk-linux-stable4-64/build/Tools/Scripts/webkitpy/webdriver_tests/webdriver_selenium_executor.py", line 57, in __init__ do_delayed_imports() File "/home/buildbot/webkitgtk/gtk-linux-stable4-64/build/Tools/Scripts/webkitpy/webdriver_tests/webdriver_selenium_executor.py", line 35, in do_delayed_imports import webkitpy.webdriver_tests.pytest_runner as pytest_runner File "/home/buildbot/webkitgtk/gtk-linux-stable4-64/build/Tools/Scripts/webkitpy/webdriver_tests/pytest_runner.py", line 32, in <module> import webkitpy.thirdparty.autoinstalled.pytest ImportError: No module named pytest ``` 1506592 1 362119 psaavedra 2019-02-15 09:25:10 -0800 Created attachment 362119 patch 1506645 2 362119 mcatanzaro 2019-02-15 10:52:20 -0800 Comment on attachment 362119 patch View in context: https://bugs.webkit.org/attachment.cgi?id=362119&action=review > Tools/Scripts/webkitpy/thirdparty/__init__.py:162 > self._install("https://files.pythonhosted.org/packages/a2/ec/415d0cccc1ed41cd7fdf69ad989da16a8d13057996371004cab4bafc48f3/pytest-3.6.2.tar.gz", > "pytest-3.6.2/src/_pytest") > + self._install("https://files.pythonhosted.org/packages/a2/ec/415d0cccc1ed41cd7fdf69ad989da16a8d13057996371004cab4bafc48f3/pytest-3.6.2.tar.gz", > + "pytest-3.6.2/src/pytest.py") I don't understand. Now it's listed twice, and the only difference is the missing underscore. Is this really the right way to do it? 1506652 3 psaavedra 2019-02-15 10:57:54 -0800 The first one is a dir (_pytest). The line added is a file (pytest.py) 1506661 4 psaavedra 2019-02-15 11:06:48 -0800 Let me work on it, I guess I could overhead the install() function to allow the definition of more than one path, dir or files. 1506668 5 ap 2019-02-15 11:08:32 -0800 My high level question is - why does it work for me? :) 1506695 6 psaavedra 2019-02-15 11:50:36 -0800 s/overhead/overload/ :-). Do you have the pytest.py module in the webkitpy/thirdparty/autoinstalled dir? If not, do you have the pytest library installed in your system? 1506925 7 ap 2019-02-15 19:37:29 -0800 $ ls webkitpy/thirdparty/autoinstalled | grep pytest _pytest pytest.py pytest_timeout.py 1506969 8 psaavedra 2019-02-15 22:18:32 -0800 I suppose that the auto-install was able to download this file at some point but not now. Once you have the pytest.py file in the autoinstalled directory you don't need to download it anymore and this error doesn't happens anymore so I guess you got this py module time ago. I'd suggest to check the change date of the pytest.py to get more information about your particular case: stat Tools/Scripts/webkitpy/thirdparty/autoinstalled/pytest.py In any case, this is a weird situation because I didn't find any commit which justify this supposition. The only that I found is this one which is where pytest is added to the auto-install but it doesn't includes the pytest.py neither: commit b6a225b5f87136ac5dec1380344c5cd082a92473 Author: carlosgc@webkit.org <carlosgc@webkit.org@268f45cc-cd09-0410-ab3c-d52691b4dbfc> Date: Sat Dec 2 14:56:26 2017 +0000 WebDriver: auto-install pytest instead of importing it from wpt tools directory https://bugs.webkit.org/show_bug.cgi?id=180243 The Tools/Scripts/webkitpy/thirdparty/autoinstalled/ directory is auto-generated by running the `Tools/Scripts/run-webdriver-tests` command and is safe to delete as it says in the README file included in the this directory. Therefore, You can reproduce the error following this instructions: rm -rf Tools/Scripts/webkitpy/thirdparty/autoinstalled/* python Tools/Scripts/run-webdriver-tests --json-output=webdriver_tests.json --release --gtk *** *** Warning: jhbuild environment not present and not running in flatpak. *** Run update-webkitgtk-libs or update-webkitgtk-flatpak before build-webkit to ensure proper testing.. *** Using port gtk Using port gtk Test configuration: <, x86, release> Test configuration: <, x86, release> Using display server xvfb Using display server xvfb Using driver at /home/psaavedra/local/workshop/debian-armhf/WebKit/WebKitBuild/Release/bin/WebKitWebDriver Using driver at /home/psaavedra/local/workshop/debian-armhf/WebKit/WebKitBuild/Release/bin/WebKitWebDriver Browser: MiniBrowser Browser: MiniBrowser Parsing expectations Parsing expectations Auto-installing package: py Auto-installing package: py Auto-installing package: pluggy Auto-installing package: pluggy Auto-installing package: more_itertools Auto-installing package: more_itertools Auto-installing package: six.py Auto-installing package: six.py Auto-installing package: atomicwrites Auto-installing package: atomicwrites Auto-installing package: funcsigs Auto-installing package: funcsigs Auto-installing package: attr Auto-installing package: attr Auto-installing package: _pytest Auto-installing package: _pytest Traceback (most recent call last): File "./Tools/Scripts/run-webdriver-tests", line 86, in <module> runner.run(args) File "/home/psaavedra/local/workshop/debian-armhf/WebKit/Tools/Scripts/webkitpy/webdriver_tests/webdriver_test_runner.py", line 70, in run runner_tests = [runner.collect_tests(tests) for runner in self._runners] File "/home/psaavedra/local/workshop/debian-armhf/WebKit/Tools/Scripts/webkitpy/webdriver_tests/webdriver_test_runner_selenium.py", line 51, in collect_tests executor = WebDriverSeleniumExecutor(self._driver, self._env) File "/home/psaavedra/local/workshop/debian-armhf/WebKit/Tools/Scripts/webkitpy/webdriver_tests/webdriver_selenium_executor.py", line 57, in __init__ do_delayed_imports() File "/home/psaavedra/local/workshop/debian-armhf/WebKit/Tools/Scripts/webkitpy/webdriver_tests/webdriver_selenium_executor.py", line 35, in do_delayed_imports import webkitpy.webdriver_tests.pytest_runner as pytest_runner File "/home/psaavedra/local/workshop/debian-armhf/WebKit/Tools/Scripts/webkitpy/webdriver_tests/pytest_runner.py", line 32, in <module> import webkitpy.thirdparty.autoinstalled.pytest ImportError: No module named pytest Once you apply my patch: ./Tools/Scripts/run-webdriver-tests --json-output=webdriver_tests.json --release --gtk *** *** Warning: jhbuild environment not present and not running in flatpak. *** Run update-webkitgtk-libs or update-webkitgtk-flatpak before build-webkit to ensure proper testing.. *** Using port gtk Using port gtk Test configuration: <, x86, release> Test configuration: <, x86, release> Using display server xvfb Using display server xvfb Using driver at /home/psaavedra/local/workshop/debian-armhf/WebKit/WebKitBuild/Release/bin/WebKitWebDriver Using driver at /home/psaavedra/local/workshop/debian-armhf/WebKit/WebKitBuild/Release/bin/WebKitWebDriver Browser: MiniBrowser Browser: MiniBrowser Parsing expectations Parsing expectations Auto-installing package: pytest.py Auto-installing package: pytest.py Auto-installing package: pytest_timeout.py Auto-installing package: pytest_timeout.py 1506975 9 ap 2019-02-15 23:09:50 -0800 The line with pytest.py was removed in r239465. Jonathan, do you remember why? 1506993 10 psaavedra 2019-02-16 01:29:43 -0800 (In reply to Alexey Proskuryakov from comment #9) > The line with pytest.py was removed in r239465. Jonathan, do you remember > why? Good catch. 1506997 11 cgarcia 2019-02-16 01:54:16 -0800 (In reply to Alexey Proskuryakov from comment #9) > The line with pytest.py was removed in r239465. Jonathan, do you remember > why? He probably thought it was duplicated by mistake, we need to bring it back. 1506999 12 362119 cgarcia 2019-02-16 02:00:31 -0800 Comment on attachment 362119 patch View in context: https://bugs.webkit.org/attachment.cgi?id=362119&action=review cq- for now just in case there was a good reason to remove that. Please, set cq+ if it was removed by mistake. >> Tools/Scripts/webkitpy/thirdparty/__init__.py:162 >> + "pytest-3.6.2/src/pytest.py") > > I don't understand. Now it's listed twice, and the only difference is the missing underscore. Is this really the right way to do it? This is simple enough, and we are not really downloading twice. 1507641 13 362119 commit-queue 2019-02-19 02:22:19 -0800 Comment on attachment 362119 patch Clearing flags on attachment: 362119 Committed r241759: <https://trac.webkit.org/changeset/241759> 1507642 14 commit-queue 2019-02-19 02:22:21 -0800 All reviewed patches have been landed. Closing bug. 1507643 15 webkit-bug-importer 2019-02-19 02:23:22 -0800 <rdar://problem/48194758> 1507668 16 jbedard 2019-02-19 08:31:30 -0800 (In reply to Carlos Garcia Campos from comment #11) > (In reply to Alexey Proskuryakov from comment #9) > > The line with pytest.py was removed in r239465. Jonathan, do you remember > > why? > > He probably thought it was duplicated by mistake, we need to bring it back. I can confirm that I thought it was duplicated by mistake. I had to reconstruct all these URLs because the old ones were invalid, it works for existing machines because if the library is already there, none of this install code gets triggered. Interesting, however, that we require this module for web driver tests. Do we run those anywhere in buildbot? We must not, I would have expected infrastructure to catch this weeks ago. 1507675 17 mcatanzaro 2019-02-19 08:52:03 -0800 Only GTK port runs WebDriver tests :( 1507677 18 jbedard 2019-02-19 08:54:41 -0800 (In reply to Michael Catanzaro from comment #17) > Only GTK port runs WebDriver tests :( Ah, that explains it. And we probably haven't deployed new GTK bots recently, hence Pablo being the first one to stumble upon this. 1508471 19 psaavedra 2019-02-21 01:20:02 -0800 (In reply to Jonathan Bedard from comment #18) > (In reply to Michael Catanzaro from comment #17) > > Only GTK port runs WebDriver tests :( > > Ah, that explains it. And we probably haven't deployed new GTK bots > recently, hence Pablo being the first one to stumble upon this. Just I idea, not a strong opinion, but maybe could be better clean up the autoinstall dir as part of the steps of the bot to avoid this kind of situations. However, it has counterparts in the case of temporary "remote resource unavailable" situations so not sure if it worst. 1508528 20 jbedard 2019-02-21 08:41:16 -0800 (In reply to Pablo Saavedra from comment #19) > (In reply to Jonathan Bedard from comment #18) > > (In reply to Michael Catanzaro from comment #17) > > > Only GTK port runs WebDriver tests :( > > > > Ah, that explains it. And we probably haven't deployed new GTK bots > > recently, hence Pablo being the first one to stumble upon this. > > Just I idea, not a strong opinion, but maybe could be better clean up the > autoinstall dir as part of the steps of the bot to avoid this kind of > situations. > > However, it has counterparts in the case of temporary "remote resource > unavailable" situations so not sure if it worst. I've discussed this code with a number of apple folks, really, we should be using virtualenv here. If we were going to make any large changes, it should be in that direction. 1509752 21 bburg 2019-02-25 12:54:05 -0800 (In reply to Jonathan Bedard from comment #20) > (In reply to Pablo Saavedra from comment #19) > > (In reply to Jonathan Bedard from comment #18) > > > (In reply to Michael Catanzaro from comment #17) > > > > Only GTK port runs WebDriver tests :( > > > > > > Ah, that explains it. And we probably haven't deployed new GTK bots > > > recently, hence Pablo being the first one to stumble upon this. > > > > Just I idea, not a strong opinion, but maybe could be better clean up the > > autoinstall dir as part of the steps of the bot to avoid this kind of > > situations. > > > > However, it has counterparts in the case of temporary "remote resource > > unavailable" situations so not sure if it worst. > > I've discussed this code with a number of apple folks, really, we should be > using virtualenv here. If we were going to make any large changes, it should > be in that direction. +1. This is especially important for WebDriver and other tests where WebKit's version of WPT tests, or the driver functionality itself, can be significantly ahead or behind the client library that's available via PyPI. Virtualenv makes it pretty easy to manage situations like this. It's used at Apple for our internal safaridriver tests. 362119 2019-02-15 09:25:10 -0800 2019-02-19 02:22:19 -0800 patch bug-194707-20190215182509.patch text/plain 1629 psaavedra U3VidmVyc2lvbiBSZXZpc2lvbjogMjQxNTg5CmRpZmYgLS1naXQgYS9Ub29scy9DaGFuZ2VMb2cg Yi9Ub29scy9DaGFuZ2VMb2cKaW5kZXggNjYxMTVhYmRlNDU1ODAxOWQ2ZjhkMzQzNDc0MmJjM2Jk MzAwNTZhZC4uYTNjZWMyYjUxYjdlNDBlNjgwN2M2ZjdjZDJmNzkyODQ1YTgwNTNmYiAxMDA2NDQK LS0tIGEvVG9vbHMvQ2hhbmdlTG9nCisrKyBiL1Rvb2xzL0NoYW5nZUxvZwpAQCAtMSwzICsxLDEz IEBACisyMDE5LTAyLTE1ICBQYWJsbyBTYWF2ZWRyYSAgPHBzYWF2ZWRyYUBpZ2FsaWEuY29tPgor CisgICAgICAgIHB5dGVzdCBpcyBub3QgY29ycmVjdGx5IGF1dG8taW5zdGFsbGVkCisgICAgICAg IGh0dHBzOi8vYnVncy53ZWJraXQub3JnL3Nob3dfYnVnLmNnaT9pZD0xOTQ3MDcKKworICAgICAg ICBSZXZpZXdlZCBieSBOT0JPRFkgKE9PUFMhKS4KKworICAgICAgICAqIFNjcmlwdHMvd2Via2l0 cHkvdGhpcmRwYXJ0eS9fX2luaXRfXy5weToKKyAgICAgICAgKEF1dG9pbnN0YWxsSW1wb3J0SG9v ay5faW5zdGFsbF9weXRlc3QpOgorCiAyMDE5LTAyLTEwICBEYXJpbiBBZGxlciAgPGRhcmluQGFw cGxlLmNvbT4KIAogICAgICAgICBSZXBsYWNlIG1vcmUgdXNlcyBvZiBTdHJpbmc6OmZvcm1hdCB3 aXRoIFN0cmluZ0NvbmNhdGVuYXRlIChtb3N0bHkgbm9uLUFwcGxlIHBsYXRmb3JtLXNwZWNpZmlj IGNhc2VzKQpkaWZmIC0tZ2l0IGEvVG9vbHMvU2NyaXB0cy93ZWJraXRweS90aGlyZHBhcnR5L19f aW5pdF9fLnB5IGIvVG9vbHMvU2NyaXB0cy93ZWJraXRweS90aGlyZHBhcnR5L19faW5pdF9fLnB5 CmluZGV4IDZiYmU1OTA4YzVkNGQ1ZTE2ZmI3YzJmOGUwZjY4MGY0YjMwOGRmM2UuLjJjODIwMzZk MmE0ODA3MTFlYWQ3MDQyNzllNzAzNWI0M2I2NmQ2ZDcgMTAwNjQ0Ci0tLSBhL1Rvb2xzL1Njcmlw dHMvd2Via2l0cHkvdGhpcmRwYXJ0eS9fX2luaXRfXy5weQorKysgYi9Ub29scy9TY3JpcHRzL3dl YmtpdHB5L3RoaXJkcGFydHkvX19pbml0X18ucHkKQEAgLTE1OCw2ICsxNTgsOCBAQCBjbGFzcyBB dXRvaW5zdGFsbEltcG9ydEhvb2sob2JqZWN0KToKICAgICAgICAgICAgICAgICAgICAgICAgICAg ICAgICJhdHRycy0xOC4xLjAvc3JjL2F0dHIiKQogICAgICAgICBzZWxmLl9pbnN0YWxsKCJodHRw czovL2ZpbGVzLnB5dGhvbmhvc3RlZC5vcmcvcGFja2FnZXMvYTIvZWMvNDE1ZDBjY2NjMWVkNDFj ZDdmZGY2OWFkOTg5ZGExNmE4ZDEzMDU3OTk2MzcxMDA0Y2FiNGJhZmM0OGYzL3B5dGVzdC0zLjYu Mi50YXIuZ3oiLAogICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgInB5dGVzdC0zLjYuMi9z cmMvX3B5dGVzdCIpCisgICAgICAgIHNlbGYuX2luc3RhbGwoImh0dHBzOi8vZmlsZXMucHl0aG9u aG9zdGVkLm9yZy9wYWNrYWdlcy9hMi9lYy80MTVkMGNjY2MxZWQ0MWNkN2ZkZjY5YWQ5ODlkYTE2 YThkMTMwNTc5OTYzNzEwMDRjYWI0YmFmYzQ4ZjMvcHl0ZXN0LTMuNi4yLnRhci5neiIsCisgICAg ICAgICAgICAgICAgICAgICAgICAgICAgICAicHl0ZXN0LTMuNi4yL3NyYy9weXRlc3QucHkiKQog CiAgICAgZGVmIF9pbnN0YWxsX3B5bGludChzZWxmKToKICAgICAgICAgc2VsZi5fZW5zdXJlX2F1 dG9pbnN0YWxsZWRfZGlyX2lzX2luX3N5c19wYXRoKCkK