aboutsummaryrefslogtreecommitdiffstats
path: root/ulri
Commit message (Collapse)AuthorAgeFilesLines
* Simplify failure detection logic and fix cleanupPascal Terjan2017-02-061-22/+19
| | | | | | | | Especially: - fetch logs when dependencies can not be installed - properly cleanup logs on the build machine in all cases - really discard other architectures if a mandatory one had already failed
* Extract get_target_arch helperOlivier Blin2016-11-241-4/+3
|
* Delete src.rpm after last non mandatory arch failsPascal Terjan2016-11-241-2/+2
|
* Record failed build in the in-memory treePascal Terjan2016-11-241-1/+2
|
* Have consistent ordering for queue statusPascal Terjan2016-11-081-1/+1
|
* Only mandatory arches are fatalPascal Terjan2016-10-201-3/+2
| | | | | Currently any previous failures casues to force fail anything succeeding later.
* Only build noarch packages oncePascal Terjan2016-10-141-0/+1
|
* Do not rebuild noarch packages that are already building/builtPascal Terjan2016-10-121-3/+8
|
* Fix reuse of {arch} part of the treePascal Terjan2016-10-111-1/+3
| | | | | | | | It is used by ulri for both builds in progress and already done, while emi expects it to be only for done ones. This causes to upload only part of mandatory arches and later reject the other.
* Fix call to build_package0.6.26Pascal Terjan2016-10-041-1/+1
|
* Reorganize code detecting build statusPascal Terjan2016-10-041-36/+34
| | | | | | | If we can't get status, log a warning and go on. If status is not ready yet, verify the build is still in progress and there is no timeout. Else, analyze the status.
* Do not give up on a build when failing to check its statusPascal Terjan2016-10-031-69/+67
|
* Remove some useless connection to the build nodePascal Terjan2016-10-031-1/+1
| | | | | If we know the build is completed there is no point checking the state of the remote process.
* Fix exclusion warning to not log 1Pascal Terjan2016-05-251-1/+1
|
* Do not consider iurt is no longer running when failing to sshPascal Terjan2016-05-211-2/+3
|
* Do not retry when a build succeeds while it is already being retriedPascal Terjan2016-05-211-2/+2
|
* Enable ServerAliveInterval by defaultPascal Terjan2016-04-151-1/+1
|
* Do not try to build failed packagesPascal Terjan2016-02-151-2/+3
|
* Move something invariant out of a loopPascal Terjan2016-02-151-8/+5
|
* Better modules splitPascal Terjan2016-02-061-9/+2
|
* Start splitting ulriPascal Terjan2016-02-061-198/+22
|
* Cleanup lines with only whitespacePascal Terjan2016-02-061-42/+42
|
* Move exclude_machine to a better placePascal Terjan2016-02-031-8/+8
|
* Use a shared function to load the state of upload treePascal Terjan2016-01-231-93/+6
|
* Do not modify $run in the tree callbacksPascal Terjan2016-01-231-3/+11
|
* Modify one less global variable in the tree callbacksPascal Terjan2016-01-231-9/+14
|
* Do not store the build bot architecture for noarch packagesPascal Terjan2016-01-231-2/+1
|
* Store packages for which the build is in progress or done in pkg_treePascal Terjan2016-01-231-13/+12
|
* Reduce the use of global variablesPascal Terjan2016-01-221-12/+6
|
* Reduce ssh timeoutPascal Terjan2015-12-261-1/+1
|
* when failing to send a build to a machine, exclude that machinePascal Terjan2015-12-261-2/+16
|
* ulri: Keep other builds going when a failure occurs on non-mandatory archOlivier Blin2015-11-171-2/+2
|
* Fix again previously fixed bug, broken againPascal Terjan2014-06-021-1/+2
|
* Only register a package as building after we managed to contact the build nodePascal Terjan2014-05-051-7/+6
|
* Set path properly when there is a lock file but no src.rpm leftPascal Terjan2014-04-301-0/+3
|
* fix indentationThierry Vignaud2014-04-201-1/+1
|
* simplify using cat*()Thierry Vignaud2014-04-191-8/+2
|
* perl_checker cleanupsThierry Vignaud2014-04-191-5/+4
|
* Fix syntax when there are dependenciesPascal Terjan2014-04-181-1/+4
|
* Fix parsing of .deps pathPascal Terjan2014-04-181-3/+2
|
* Add a way to have a build waiting for others.Pascal Terjan2014-04-181-0/+47
| | | | | | | | | | | | | | | | This is a list of prefix, so if one of the builds fails dependent ones will be rejected. It would be nicer for a user perspective to specify that we want to wait for the src.rpm of foo >= XX.YY to be available, with some time limit, as it would get built if the original build fails but get submitted again, but that would be much more complicated to implement. This simple solution allows a Youri::Submit action to submit build of dependant packages while the upload is not finished yet. For example this will allow to automatically rebuild kmod-* and drakx-installer-images after a new kernel is uploaded.
* Update defaults and kill obsolete mappingsPascal Terjan2014-01-191-36/+55
|
* Try to improve cleanup (and not forget a build when another arch fails)Pascal Terjan2014-01-191-40/+9
|
* Record built packages to allow cleanup when only one arch failsPascal Terjan2014-01-191-1/+3
|
* Still cleanup the queues when fetching logs for a failure fail (Probably ↵Pascal Terjan2014-01-191-7/+2
| | | | never actually happens)
* Fix ulriPascal Terjan2013-01-111-1/+1
|
* Do not mark an arch done until all files are copiedPascal Terjan2013-01-111-6/+9
|
* fix tooThierry Vignaud2013-01-011-1/+1
|
* simplify like emi wasThierry Vignaud2012-12-111-8/+2
|
* perl_checker cleanupThierry Vignaud2012-12-111-5/+5
|