Community
Participate
Working Groups
We have a very old set of Browser(1-8) tests, which were written before the days of jUnit and have some problems. It's time to update them and merge them into test_*_Browser instead. Problems: - They usually combine multiple tests into a single one big, so if something doesn't work, it's not clear what's broken. - They are unnecessarily slow. E.g even when a condition is reached, they still wait on a hard timeout and only check for condition after. - The wait mechanism is problematic, in that the threading can cause unexpected behaviour. e.g Bug webkitBrowser1Crash ( Bug 515471 ) The wait logic in test_*_browser is better. I shall be submitting relevant patches.
New Gerrit change created: https://git.eclipse.org/r/97997
New Gerrit change created: https://git.eclipse.org/r/98075
New Gerrit change created: https://git.eclipse.org/r/98088
New Gerrit change created: https://git.eclipse.org/r/98260
New Gerrit change created: https://git.eclipse.org/r/98389
New Gerrit change created: https://git.eclipse.org/r/98572
New Gerrit change created: https://git.eclipse.org/r/98955
New Gerrit change created: https://git.eclipse.org/r/98960
I want to verify that tests work well on Cocoa/Win32, then I'll merge the new jUnits.
(In reply to Leo Ufimtsev from comment #9) > I want to verify that tests work well on Cocoa/Win32, then I'll merge the > new jUnits. Eta: next week-ish.
I'm setting up Cocoa/Win32 servers to test these patches prior to merge. Will get to this after.
Gerrit change https://git.eclipse.org/r/97997 was merged to [master]. Commit: http://git.eclipse.org/c/platform/eclipse.platform.swt.git/commit/?id=4edf9c6d152dd35087da29bd12ec4bacd3d13d21
Gerrit change https://git.eclipse.org/r/98075 was merged to [master]. Commit: http://git.eclipse.org/c/platform/eclipse.platform.swt.git/commit/?id=902a410605c4c3ebba72480b3628d720948a5e18
All outstanding patches fixed/updated & tested on Win10/OSX/Linux. Now waiting for master to re-open and merge.
Gerrit change https://git.eclipse.org/r/98088 was merged to [master]. Commit: http://git.eclipse.org/c/platform/eclipse.platform.swt.git/commit/?id=8ded377259b12f1786b4be69aef07f5b734a960b
Gerrit change https://git.eclipse.org/r/98260 was merged to [master]. Commit: http://git.eclipse.org/c/platform/eclipse.platform.swt.git/commit/?id=a91d746a562f35b22ae5cb30e2f79ff706a16eb6
Gerrit change https://git.eclipse.org/r/98389 was merged to [master]. Commit: http://git.eclipse.org/c/platform/eclipse.platform.swt.git/commit/?id=f3feb49d109d6bb2bdd90deca5ed6b519f351900
Gerrit change https://git.eclipse.org/r/98572 was merged to [master]. Commit: http://git.eclipse.org/c/platform/eclipse.platform.swt.git/commit/?id=b8e4a78fa923dbb19095e7f76c123fd14c957c8f
Gerrit change https://git.eclipse.org/r/98955 was merged to [master]. Commit: http://git.eclipse.org/c/platform/eclipse.platform.swt.git/commit/?id=79b7f3a6162b4fc156d848feb7d2f5bc63b052a3
Gerrit change https://git.eclipse.org/r/98960 was merged to [master]. Commit: http://git.eclipse.org/c/platform/eclipse.platform.swt.git/commit/?id=bb00ce6545cb194e0c1434a49bfcc0f2b89d562f
All old Browser_test* cases were replaced by improved Test_*_Browser tests that run faster and break large tests apart into smaller chunks.
New Gerrit change created: https://git.eclipse.org/r/102812
Gerrit change https://git.eclipse.org/r/102812 was merged to [master]. Commit: http://git.eclipse.org/c/platform/eclipse.platform.swt.git/commit/?id=3b484faff52cb8ef82cf78bf5a78f0a452f8be3f
How can you mark a bug as VERFIFIED when you not even waited for the first official test results (http://download.eclipse.org/eclipse/downloads/drops4/I20170809-2000/testResults.php)? ;-) Test_org_eclipse_swt_browser_Browser#test_setUrl_local so far failed on two platforms but with different errors. Let's see what the other platforms will say.
(In reply to Dani Megert from comment #24) > How can you mark a bug as VERFIFIED when you not even waited for the first > official test results > (http://download.eclipse.org/eclipse/downloads/drops4/I20170809-2000/ > testResults.php)? ;-) > > Test_org_eclipse_swt_browser_Browser#test_setUrl_local so far failed on two > platforms but with different errors. Let's see what the other platforms will > say. That's a good point. Local maven tests worked on linux/cocoa/win. But I should have waited to verify it worked on servers. I'll check.
New Gerrit change created: https://git.eclipse.org/r/102896
Gerrit change https://git.eclipse.org/r/102896 was merged to [master]. Commit: http://git.eclipse.org/c/platform/eclipse.platform.swt.git/commit/?id=1a824dc95731ccc896ed8db58ce51604b2acd78c
(In reply to Eclipse Genie from comment #26) > New Gerrit change created: https://git.eclipse.org/r/102896 Should hopefully fix the test_setUrl_local() issue. I can't reproduce it on my local linux/mac/win10 machines, but have a theory as to why they fail. I'll check builds tomorrow for progress.
test_setUrl_local() now passed on Cocoa. (There is a graphics/setCursor test case failing). On windows test_setUrl_remote() failed due to a timeout. (before fix it said "page could not be displayed". Issue could have occurred due to bad internet connection at server or slow performance. I will check again on monday/tuesday. If test continues to fail I'll investigate further.
The newly added "test_setUrl_remote()" test doesn't seem to be stable. 2/3 times it worked, but yesterday it again timed out on windows. It's possible that slow internet connection is to blame. I'll attempt to increase the timeout specifically for that case, to see if there will be an improvement.
New Gerrit change created: https://git.eclipse.org/r/103069
Gerrit change https://git.eclipse.org/r/103069 was merged to [master]. Commit: http://git.eclipse.org/c/platform/eclipse.platform.swt.git/commit/?id=6ea2e317fce7d0d60e3518aa43ba05dca81dacda
Let's wait for a few days and see what happens.
New Gerrit change created: https://git.eclipse.org/r/103113
(In reply to Eclipse Genie from comment #34) > New Gerrit change created: https://git.eclipse.org/r/103113 test_setUrl_remote() still fails sporadically on timeout. Sometimes it works, sometimes it times out. This test needs a working internet, so I presume it times out if internet is not working prorerly. Adding an assumption that checks internet. If internet doesn't work, skip test. This should make this test less noisy. Note, other tests test setUrl() against local files, so setUrl is still well covered even if intent sometimes doesn't work. Will followup in a few days.
Gerrit change https://git.eclipse.org/r/103113 was merged to [master]. Commit: http://git.eclipse.org/c/platform/eclipse.platform.swt.git/commit/?id=ab84ce04cbe7f494e3746d19d40968908097cd5d
(In reply to Leo Ufimtsev from comment #35) > (In reply to Eclipse Genie from comment #34) > > New Gerrit change created: https://git.eclipse.org/r/103113 > > test_setUrl_remote() still fails sporadically on timeout. > Sometimes it works, sometimes it times out. This test needs a working > internet, so I presume it times out if internet is not working prorerly. > > Adding an assumption that checks internet. If internet doesn't work, skip > test. > This should make this test less noisy. Note, other tests test setUrl() > against local files, so setUrl is still well covered even if intent > sometimes doesn't work. > > Will followup in a few days.
(In reply to Leo Ufimtsev from comment #35) > (In reply to Eclipse Genie from comment #34) > > New Gerrit change created: https://git.eclipse.org/r/103113 > > test_setUrl_remote() Passed today, but test did take a very long time: 127.322 seconds. Will keep an eye...
New Gerrit change created: https://git.eclipse.org/r/103408
Gerrit change https://git.eclipse.org/r/103408 was merged to [master]. Commit: http://git.eclipse.org/c/platform/eclipse.platform.swt.git/commit/?id=1c87b744636da422526f92fb4597383cdb1f2aad
setText seems to sometimes fail on windows. "about:blank" seems to get loaded before the actual page. Attempt to fix by skipping "about:blank" pages: https://git.eclipse.org/r/#/c/103408/ (simmilar to how it was done in older test in the past)
No failing tests today. Will continue to monitor.
No failing browser test cases related to this refactoring observed so far. -> closing. Will followup in 1-2 weeks to verify. If someone spots something odd, please reopen/comment. Thanks.
