-
Notifications
You must be signed in to change notification settings - Fork 82
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Seperate Build systems might be overriding each others release tag #909
Comments
As well, I think that non master branch builds from Travis are also getting uploaded to latest-master tag, which would be wrong. |
Last few builds I've seen and things seem to be working normally. Closing. |
Seems to be happening again. Now I am seeing that the Core.zip is not being uploaded. |
Here is a recent log portion from latest-master Travis CI build during the binary upload section (in this case, core.zip):
It might indicate an older version of python contributing to the upload issue. But then again, the Ubuntu version is Ubuntu 12.04.5 LTS, so I am unsure if an upgrade is even possible. It might not solve the issue anyways, but thought it was good to add the possibility to the issue comments. |
Unless something new happens, I think the bug you're tracking is well known for years and easily worked around: as soon as an artifact fails building, it will not be uploaded in the latest-master release. If it is one of the first artifacts (one for appveyor, one for travis, one for the mac_travis) of the next build run is trying the upload, it will cause releases to be partial for ever because each build will first upload in the next release to complete the artifact set then will detects that artifact are being overwritten and create a new release ... One easy way to work around that is to manually re-run the build of the missing artifact like I just did with Appveyor. |
@xawotihs Thanks for the info. Yes for Travis, I have been able to restart builds that succeeded on the second pass with no code changes to work around this. For Appveyor however, I do not have that kind of access. I guess we would have to rely on you to catch the fail and spark a new build. Seems like a hassle to ask you to watch every build though. Really only affects the Windows build as it is from Appveyor. Speaking of which, whatever happened to the IOS auto build? I think it's disabled in builds now. |
|
What happened to the releases? It seems wagic qt is the only one uploaded? |
Haven't been involved for a while (new job), but have seen this is still an issue. Although I have looked in every once in a while and saw at least some more build artifacts published than just QT linux, but as always, is a roll of the dice which ones show up. Just looked at the last build log on travis, first of all, even looking at "raw" log shows colour escape codes in it. Don't think that is related in any way, but may indicate how broken that system is. Otherwise, builds compile fine, the upload messages at the end look fine except for the SSL warnings. So. I manually restarted the master branch build just on Travis, let's see what happens. |
UPDATE: |
UPDATE UPDATE: |
I have just restarted manually my commit and the results the same. Only Linux Qt is available. Does the phyton script erases previous one uploaded, becuase the last list is Linux Qt, so perhaps it erases/overwrite others? `travis_time:end:1c1829c0:start=1489368220355950251,finish=1489368261233355472,duration=40877405221 travis_time:end:07024e72:start=1489368261238777853,finish=1489368263823758258,duration=2584980405 travis_time:end:1de332ee:start=1489368263828758614,finish=1489368265495832564,duration=1667073950 travis_time:end:064e0e4c:start=1489368265504920045,finish=1489368267010882757,duration=1505962712 |
@kevlahnota This is just a guess, but possibly because the upload script is getting called a bunch of times at once, maybe there is a threading oddness, where something in python abandons previous calls and only the last one finishes. Or something along those lines. That is just a guess. Really have no idea, but want to fix this for sure. I think many people are missing out on Wagic because of this. |
OH wait!
This should be happening far earlier in the sequence. It may not be a script error, but again some kind of threading oddness. |
Actually, I was wrong. I see how the upload script goes through a loop of checking assets and if it finds a match, it deletes the tag? and because recently, only the QT artifact was there, that is when it did the tag delete. does the script block on this line before continuing? I admit, I am a python newb. Perhaps I will experiment on my fork of the project. |
@kevlahnota Well SOMETHING happened, on that last commit/Travis build, now all the release artifacts are there. Should we stamp a build for permenant release? v0.20.1 was July 2016. I am not sure of what the versioning scheme was decided upon. |
Yes but how? I have some commits left (still testing some cards) but I think to be safe, I agree on stamping the new build for permanent release. |
I saw today that the build artifacts got messed up again. I went to try another build in order to try to stamp something and this time, the core failed to upload:
|
AppVeyor (Win, IOS) build upload to latest-master might be getting overridden by Travis CI (everything else) build latest-master branch uploads because a new tag is created when each build system finishes and uploads.
The text was updated successfully, but these errors were encountered: