101261 2012-11-05 14:46:24 -0800 test-webkitpy shows failure output on my linux box 2013-01-08 11:20:56 -0800 1 1 1 Unclassified WebKit Tools / Tests 528+ (Nightly build) Unspecified Unspecified RESOLVED FIXED P2 Normal --- 1 eric zan abarth dpranke d-r peter rniwa sam webkit.review.bot zan oldest_to_newest 758920 0 eric 2012-11-05 14:46:24 -0800 % test-webkitpy Suppressing most webkitpy logging while running unit tests. Skipping tests in the following modules or packages because they are really, really, slow: webkitpy.common.checkout.scm.scm_unittest (https://bugs.webkit.org/show_bug.cgi?id=31818; use --all to include) [100/1598] webkit2.messages_unittest.HeaderTest.test_header (+15)Failure: No such entity Failure: No such entity Failure: No such entity Ran 1598 tests in 3.133s OK Unclear if it's webkit2-code related or not yet. 758976 1 dpranke 2012-11-05 15:52:03 -0800 It runs cleanly for me on my linux box synced to r133537. There were some wk2-related failures briefly last week; is it possible you're synced to one of those revisions? 799369 2 peter 2013-01-03 07:13:55 -0800 I'm seeing this as well. A Google search on the error shows that it's likely PulseAudio causing this. Are you on Precise? The test it shows the failure messages show on differs between tests. I'm not seeing it when executing "test-webkitpy -j1" either, and with -j2 it shows once. ---------- [162/1614] webkitpy.common.net.buildbot.buildbot_unittest.BuilderTest.test_failure_and_flaky passedFailure: No such entity Failure: No such entity Failure: No such entity Ran 1614 tests in 3.034s OK 799373 3 peter 2013-01-03 07:33:59 -0800 I'm seeing this consistently when running webkitpy.layout_tests.port.efl_unittest and gtk_unittest. Both ports use the PulseAudioSanitizer class to unload the "module-stream-restore" PulseAudio module before the test run (as part of setup_test_run()), and restore it again as part of clean_up_test_run(). This executes the following commands: $ pactl list short modules $ pactl unload-module [index of module-stream-restore module] .. test run .. $ pactl load-module module-stream-resource Adding a print right before the call to unload-module shows that this is being executed twice (or more), so we're running into a concurrency issue here. The class was added in http://wkrev.com/115478. 799382 4 peter 2013-01-03 07:51:39 -0800 To clarify: two or more threads are simultaneously running PulseAudioSanitizer._unload_pulseaudio_module(), all execute $ pactl list short modules ..and all see that the "module-stream-resource" module is loaded, causing them try to unload it. The first unload succeeds, subsequent unloads show the "Failure: No such entity" message. Easy solution would be to protect the call to _unload_pulseaudio_module and _restore_pulseaudio_module in a multiprocess.Lock(), but this would still mean that we're loading and unloading the module two times per webkitpy test run. Dominik, do you have any preferred solution? IIRC running failed tests again is seen as a separate run, so in case of failed layout tests the PulseAudio module would also be loaded and unloaded twice. 799583 5 dpranke 2013-01-03 12:53:53 -0800 Note that you can mark python unit tests in webkitpy as needing to be run in isolation by prepending "serial_" to the test name. I would do that rather than do locking manually. 800245 6 d-r 2013-01-04 09:11:34 -0800 (In reply to comment #4) > Dominik, do you have any preferred solution? IIRC running failed tests again is seen as a separate run, so in case of failed layout tests the PulseAudio module would also be loaded and unloaded twice. Do you think what Dirk suggests would be the right fix, or do you mean that you see this when running the unit tests but you consider it an issue in other situations, too? 800924 7 zan 2013-01-05 09:16:53 -0800 (In reply to comment #4) > Dominik, do you have any preferred solution? IIRC running failed tests again is seen as a separate run, so in case of failed layout tests the PulseAudio module would also be loaded and unloaded twice. But that isn't done concurrently, so there's no 'No such entity' output in that case (AFAIK). I'm not sure there's a way of enforcing running unit tests from two separate files separately. How about mocking out the PulseAudioSanitizer methods? 801328 8 dpranke 2013-01-07 09:12:24 -0800 (In reply to comment #7) > I'm not sure there's a way of enforcing running unit tests from two separate files separately. How about mocking out the PulseAudioSanitizer methods? The serial_ feature works across files. Of course, mocking out the routines would probably work as well. 802257 9 181697 zan 2013-01-08 08:15:35 -0800 Created attachment 181697 Patch 802267 10 peter 2013-01-08 08:32:58 -0800 (In reply to comment #5) > Note that you can mark python unit tests in webkitpy as needing to be run in isolation by prepending "serial_" to the test name. I would do that rather than do locking manually. In this case it's a few unit tests for both the EFL and GTK ports that call setup_test_run(), I'm not sure if running them serially is the right solution as it would still unload and restore the module. (In reply to comment #7) > (In reply to comment #4) > > Dominik, do you have any preferred solution? IIRC running failed tests again is seen as a separate run, so in case of failed layout tests the PulseAudio module would also be loaded and unloaded twice. > > But that isn't done concurrently, so there's no 'No such entity' output in that case (AFAIK). > > I'm not sure there's a way of enforcing running unit tests from two separate files separately. How about mocking out the PulseAudioSanitizer methods? The module is being unloaded concurrently as two or more tests are invoking setup_test_run(), and thus the PulseAudio unload method, at the same time. I think Zan's solution makes sense, and it stops the unit tests from unloading and restoring the PulseAudio module. Thank you for looking in to it -- it solves the issue for me. 802389 11 181697 eric 2013-01-08 10:59:17 -0800 Comment on attachment 181697 Patch View in context: https://bugs.webkit.org/attachment.cgi?id=181697&action=review Seems fine. Makes me think that this Port object may be doing too much. :) > Tools/Scripts/webkitpy/layout_tests/port/efl_unittest.py:43 > + port._unload_pulseaudio_module = lambda: None > + port._restore_pulseaudio_module = lambda: None I guess there is no PulseAudio object on these ports that you can just mock? The port itself handles teh loading/unloading (and presumably actions with) this code? 802402 12 181697 zan 2013-01-08 11:06:27 -0800 Comment on attachment 181697 Patch View in context: https://bugs.webkit.org/attachment.cgi?id=181697&action=review >> Tools/Scripts/webkitpy/layout_tests/port/efl_unittest.py:43 >> + port._restore_pulseaudio_module = lambda: None > > I guess there is no PulseAudio object on these ports that you can just mock? The port itself handles teh loading/unloading (and presumably actions with) this code? The port interfaces inherit from the PulseAudioSanitizer and call these methods on test run setup/cleanup. https://trac.webkit.org/browser/trunk/Tools/Scripts/webkitpy/layout_tests/port/efl.py#L37 https://trac.webkit.org/browser/trunk/Tools/Scripts/webkitpy/layout_tests/port/gtk.py#L38 Please give the final thumbs-up for landing this if all seems fine to you. 802412 13 eric 2013-01-08 11:14:49 -0800 This is fine. But eventually we should fix that. A port "is not" a PulseAudioSanitizer, it might "have a", but the usage of inheritance here is wrong IMO. :) But that's a separate bug. 802420 14 zan 2013-01-08 11:18:43 -0800 (In reply to comment #13) > But that's a separate bug. Filed bug #106354. 802421 15 181697 zan 2013-01-08 11:20:50 -0800 Comment on attachment 181697 Patch Clearing flags on attachment: 181697 Committed r139084: <http://trac.webkit.org/changeset/139084> 802422 16 zan 2013-01-08 11:20:56 -0800 All reviewed patches have been landed. Closing bug. 181697 2013-01-08 08:15:35 -0800 2013-01-08 11:20:49 -0800 Patch bug-101261-20130108171240.patch text/plain 3263 zan U3VidmVyc2lvbiBSZXZpc2lvbjogMTM5MDU2CmRpZmYgLS1naXQgYS9Ub29scy9DaGFuZ2VMb2cg Yi9Ub29scy9DaGFuZ2VMb2cKaW5kZXggNGM5NjRlZWQwMzJmMGY4OWJhYjNiYTFkYjc5YjJkNTc0 ZTFmMTg2ZC4uNzM2Mjg3YjEwYTJmNjIyYWYwMTBkNDZkYThkYzIyYzUwZDA1ZWY1MSAxMDA2NDQK LS0tIGEvVG9vbHMvQ2hhbmdlTG9nCisrKyBiL1Rvb2xzL0NoYW5nZUxvZwpAQCAtMSwzICsxLDIy IEBACisyMDEzLTAxLTA4ICBaYW4gRG9iZXJzZWsgIDx6YW5kb2JlcnNla0BnbWFpbC5jb20+CisK KyAgICAgICAgdGVzdC13ZWJraXRweSBzaG93cyBmYWlsdXJlIG91dHB1dCBvbiBteSBsaW51eCBi b3gKKyAgICAgICAgaHR0cHM6Ly9idWdzLndlYmtpdC5vcmcvc2hvd19idWcuY2dpP2lkPTEwMTI2 MQorCisgICAgICAgIFJldmlld2VkIGJ5IE5PQk9EWSAoT09QUyEpLgorCisgICAgICAgIE1vY2sg b3V0IHRoZSBQdWxzZUF1ZGlvU2FuaXRpemVyIG1ldGhvZHMgd2hlbiBjcmVhdGluZyB0ZXN0aW5n IGluc3RhbmNlcworICAgICAgICBvZiBFZmxQb3J0IGFuZCBHdGtQb3J0IGludGVyZmFjZXMgaW4g dW5pdCB0ZXN0cy4gVGhpcyBzdG9wcyB0aHJvd2luZyB1cAorICAgICAgICBQdWxzZUF1ZGlvIG91 dHB1dCB3aGVuIHRoZSBzYW5pdGl6ZXIgd2FzIHRyeWluZyB0byB1bmxvYWQgbW9kdWxlcyB3aGlj aAorICAgICAgICB3ZXJlIGFscmVhZHkgdW5sb2FkZWQgYnkgYW5vdGhlciB1bml0IHRlc3QgdGhh dCB3YXMgYmVpbmcgcnVuIGluIHBhcmFsbGVsLgorCisgICAgICAgICogU2NyaXB0cy93ZWJraXRw eS9sYXlvdXRfdGVzdHMvcG9ydC9lZmxfdW5pdHRlc3QucHk6CisgICAgICAgIChFZmxQb3J0VGVz dCk6CisgICAgICAgIChFZmxQb3J0VGVzdC5tYWtlX3BvcnQpOgorICAgICAgICAqIFNjcmlwdHMv d2Via2l0cHkvbGF5b3V0X3Rlc3RzL3BvcnQvZ3RrX3VuaXR0ZXN0LnB5OgorICAgICAgICAoR3Rr UG9ydFRlc3QpOgorICAgICAgICAoR3RrUG9ydFRlc3QubWFrZV9wb3J0KToKKwogMjAxMy0wMS0w OCAgSGVpa2tpIFBhYWphbmVuICA8aGVpa2tpLnBhYWphbmVuQHBhbG0uY29tPgogCiAgICAgICAg IFtRdF0gRml4IGxpYnhtbDIgY29uZmlnIHRlc3QKZGlmZiAtLWdpdCBhL1Rvb2xzL1NjcmlwdHMv d2Via2l0cHkvbGF5b3V0X3Rlc3RzL3BvcnQvZWZsX3VuaXR0ZXN0LnB5IGIvVG9vbHMvU2NyaXB0 cy93ZWJraXRweS9sYXlvdXRfdGVzdHMvcG9ydC9lZmxfdW5pdHRlc3QucHkKaW5kZXggMWFjNjg3 YjE4N2NkMjM5NzIxZGVmZWY4NTcyM2I3MDk5YjJhMjM4Ni4uYjRlMjFhYjZiNzRiYjQ0MTM5OGYz MmE3Zjg4YTg4MDVmYzIzZjRmZiAxMDA2NDQKLS0tIGEvVG9vbHMvU2NyaXB0cy93ZWJraXRweS9s YXlvdXRfdGVzdHMvcG9ydC9lZmxfdW5pdHRlc3QucHkKKysrIGIvVG9vbHMvU2NyaXB0cy93ZWJr aXRweS9sYXlvdXRfdGVzdHMvcG9ydC9lZmxfdW5pdHRlc3QucHkKQEAgLTM2LDYgKzM2LDEzIEBA IGNsYXNzIEVmbFBvcnRUZXN0KHBvcnRfdGVzdGNhc2UuUG9ydFRlc3RDYXNlKToKICAgICBwb3J0 X25hbWUgPSAnZWZsJwogICAgIHBvcnRfbWFrZXIgPSBFZmxQb3J0CiAKKyAgICAjIEFkZGl0aW9u YWxseSBtb2NrcyBvdXQgdGhlIFB1bHNlQXVkaW9TYW5pdGl6ZXIgbWV0aG9kcy4KKyAgICBkZWYg bWFrZV9wb3J0KHNlbGYsIGhvc3Q9Tm9uZSwgcG9ydF9uYW1lPU5vbmUsIG9wdGlvbnM9Tm9uZSwg b3NfbmFtZT1Ob25lLCBvc192ZXJzaW9uPU5vbmUsICoqa3dhcmdzKToKKyAgICAgICAgcG9ydCA9 IHN1cGVyKEVmbFBvcnRUZXN0LCBzZWxmKS5tYWtlX3BvcnQoaG9zdCwgcG9ydF9uYW1lLCBvcHRp b25zLCBvc19uYW1lLCBvc192ZXJzaW9uLCAqKmt3YXJncykKKyAgICAgICAgcG9ydC5fdW5sb2Fk X3B1bHNlYXVkaW9fbW9kdWxlID0gbGFtYmRhOiBOb25lCisgICAgICAgIHBvcnQuX3Jlc3RvcmVf cHVsc2VhdWRpb19tb2R1bGUgPSBsYW1iZGE6IE5vbmUKKyAgICAgICAgcmV0dXJuIHBvcnQKKwog ICAgIGRlZiB0ZXN0X3Nob3dfcmVzdWx0c19odG1sX2ZpbGUoc2VsZik6CiAgICAgICAgIHBvcnQg PSBzZWxmLm1ha2VfcG9ydCgpCiAgICAgICAgIHBvcnQuX2V4ZWN1dGl2ZSA9IE1vY2tFeGVjdXRp dmUoc2hvdWxkX2xvZz1UcnVlKQpkaWZmIC0tZ2l0IGEvVG9vbHMvU2NyaXB0cy93ZWJraXRweS9s YXlvdXRfdGVzdHMvcG9ydC9ndGtfdW5pdHRlc3QucHkgYi9Ub29scy9TY3JpcHRzL3dlYmtpdHB5 L2xheW91dF90ZXN0cy9wb3J0L2d0a191bml0dGVzdC5weQppbmRleCA3NWQ5MTVmZjMyMGNlNzJl MWVhYjdmOTEzMjYzM2U3MjY3ZDdlNGUyLi5hOWZhYjM1ODdiZTVlNDliNTg4ZWUxODIzZmQ3Y2M2 OWY0YzFkYjczIDEwMDY0NAotLS0gYS9Ub29scy9TY3JpcHRzL3dlYmtpdHB5L2xheW91dF90ZXN0 cy9wb3J0L2d0a191bml0dGVzdC5weQorKysgYi9Ub29scy9TY3JpcHRzL3dlYmtpdHB5L2xheW91 dF90ZXN0cy9wb3J0L2d0a191bml0dGVzdC5weQpAQCAtNDMsNiArNDMsMTMgQEAgY2xhc3MgR3Rr UG9ydFRlc3QocG9ydF90ZXN0Y2FzZS5Qb3J0VGVzdENhc2UpOgogICAgIHBvcnRfbmFtZSA9ICdn dGsnCiAgICAgcG9ydF9tYWtlciA9IEd0a1BvcnQKIAorICAgICMgQWRkaXRpb25hbGx5IG1vY2tz IG91dCB0aGUgUHVsc2VBdWRpb1Nhbml0aXplciBtZXRob2RzLgorICAgIGRlZiBtYWtlX3BvcnQo c2VsZiwgaG9zdD1Ob25lLCBwb3J0X25hbWU9Tm9uZSwgb3B0aW9ucz1Ob25lLCBvc19uYW1lPU5v bmUsIG9zX3ZlcnNpb249Tm9uZSwgKiprd2FyZ3MpOgorICAgICAgICBwb3J0ID0gc3VwZXIoR3Rr UG9ydFRlc3QsIHNlbGYpLm1ha2VfcG9ydChob3N0LCBwb3J0X25hbWUsIG9wdGlvbnMsIG9zX25h bWUsIG9zX3ZlcnNpb24sICoqa3dhcmdzKQorICAgICAgICBwb3J0Ll91bmxvYWRfcHVsc2VhdWRp b19tb2R1bGUgPSBsYW1iZGE6IE5vbmUKKyAgICAgICAgcG9ydC5fcmVzdG9yZV9wdWxzZWF1ZGlv X21vZHVsZSA9IGxhbWJkYTogTm9uZQorICAgICAgICByZXR1cm4gcG9ydAorCiAgICAgZGVmIHRl c3RfZXhwZWN0YXRpb25zX2ZpbGVzKHNlbGYpOgogICAgICAgICBwb3J0ID0gc2VsZi5tYWtlX3Bv cnQoKQogICAgICAgICBzZWxmLmFzc2VydEVxdWFscyhwb3J0LmV4cGVjdGF0aW9uc19maWxlcygp LCBbJy9tb2NrLWNoZWNrb3V0L0xheW91dFRlc3RzL3BsYXRmb3JtL2d0ay9UZXN0RXhwZWN0YXRp b25zJywgJy9tb2NrLWNoZWNrb3V0L0xheW91dFRlc3RzL3BsYXRmb3JtL2d0ay13azEvVGVzdEV4 cGVjdGF0aW9ucyddKQo=