Problem
Regressions in I-Builds seem to be ignored and unnoticed by most committers.
Suggestion
When an I-Build has regression compared to some baselines (eg previous I-Build, previous release...), send a mail to the mailing-list one that matter.
This could most likely be achieved as part of each platform-specific build. The tricky part seems to be mostly getting the JUnit reports in XML format. once we have such file, the list of regressions can be sorted out with some bash, eg
$ curl --compressed https://ci.eclipse.org/ls/job/jdt-core-incubator/job/dom-with-javac/lastSuccessfulBuild/testReport/api/xml | xmllint --xpath '/testResult/suite/case[status != "PASSED"]/*[name() = "className" or name() = "name"]/text()' - | sed 'N; s/\n/ /' - | sort > baselineFailures
$ curl --compressed https://ci.eclipse.org/ls/job/jdt-core-incubator/job/PR-714/lastSuccessfulBuild/testReport/api/xml | xmllint --xpath '/testResult/suite/case[status != "PASSED"]/*[name() = "className" or name() = "name"]/text()' - | sed 'N; s/\n/ /' - | sort > prFailures
$ comm -2 -3 prFailures baselineFailures > regressions
when regressions is not an empty file, a mail could then be sent with the list.
Community
Problem
Regressions in I-Builds seem to be ignored and unnoticed by most committers.
Suggestion
When an I-Build has regression compared to some baselines (eg previous I-Build, previous release...), send a mail to the mailing-list one that matter.
This could most likely be achieved as part of each platform-specific build. The tricky part seems to be mostly getting the JUnit reports in XML format. once we have such file, the list of regressions can be sorted out with some bash, eg
when
regressionsis not an empty file, a mail could then be sent with the list.Community