Repository navigation
Tests randomly failing in AppVeyor #316
Description
Regarding #310 and #315,
AppVeyor just randomly fails partway through testing on v10 with no output or reasoning. Try to log any output and it starts working, so it's seemingly impossible to figure out for now. Gonna hit up @FeodorFitsner to see if he has any ideas. A few times now the same commit has given different results (eg make commit A, get result a, pushing commit B on top of A, get result b, and then force push commit A back again as head, and get a different result c.
Activity
FeodorFitsner commentedon Dec 5, 2014
Could you drop the links to AV build jobs in questions?
maxkorp commentedon Dec 5, 2014
Sure thing!
https://ci.appveyor.com/project/TimBranyen/nodegit/build/793
https://ci.appveyor.com/project/maxkorp/nodegit/build/7
Those point to the same commit, but if you look at the console output one gets through one more test than the other. Multiple commits with no affecting changes (eg logging a static string) were sometimes yielding different results (eg a success, or different numbers of tests passing first). At one point, the branch passed, i made a commit which made it fail, then I pushed another branch at the same commit as was passing, then that failed. It's all odd, and seems to have started last night.
johnhaley81 commentedon Jan 5, 2015
Is this fixed via #324?
johnhaley81 commentedon Jan 7, 2015
Nope :(
Still an issue.
johnhaley81 commentedon Jan 15, 2015
Trying to ferret the issue out via #352.
It seems the be that the async cloning is killing something somewhere. I was able to repo it once locally and only through Powershell and not cmd.exe or cmder.
Removing just one of the clone tests seems to fix the issue but that's a pretty crappy solution. Any help with this though would be great.
johnhaley81 commentedon Jan 16, 2015
So we decided that for the time being we're going to skip some of the clone tests and get appveyor passing. It's not the best solution but we've spent a lot of time trying to figure out what is actually the issue and it just seems like a timing thing. After #352 is merged I'll close this out.
johnhaley81 commentedon Jan 16, 2015
Hey @FeodorFitsner, is there an environment variable that I can tie into to know that code is being run on AppVeyor?
Alternatively you can help me figure out why the clone tests are dying. :)
johnhaley81 commentedon Jan 16, 2015
10 seconds after I ask I find http://www.appveyor.com/docs/environment-variables
johnhaley81 commentedon Jan 20, 2015
#352 was finalized and we decided to remove node v0.12/0.11 from the test matrix for AppVeyor for the time being since we now support io.js. AppVeyor is passing and now good so I think we can close this issue out now.
maxkorp commentedon Feb 4, 2015
This issue is very much alive. We still get it in the clone tests, and I suspect thats what is happening with the status tests in #374.
@FeodorFitsner, any chance you can take a look at this?
FeodorFitsner commentedon Feb 4, 2015
Have you tried updating npm to 2.0: http://www.appveyor.com/docs/lang/nodejs-iojs#locking-errors
maxkorp commentedon Feb 4, 2015
No, but that doesnt quite seem like the issue. I think the problem may have been something on our side related to slashes. Path module generally handles them, but there were some slipping through in a string passed to fs.exec. Fixing all of that up right now to see if it helps.
maxkorp commentedon Feb 4, 2015
Hrm, this is finally occurring outside of appveyor, and happening locally on windows.
maxkorp commentedon Feb 10, 2015
Figured it out. A few weird synchronicity issues, amongst other things. Fixed here:
#388