Open TomasTomecek opened 1 year ago
IMHO, we should not need to go to machine learning, plain simple highlighting of the error lines would make a huge improvement. This can be achieve with at least one of the next two steps:
PS. Try to not over-engineer the coloring, find the easiest way to enable it.
Using regexes to highlight the lines with errors, in the end with less than 10 regexes we could probably highlight he important lines from >99% of the logs.
This is an interesting idea. There's a similar Mock approach https://github.com/rpm-software-management/mock/issues/279
If we tweaked the tool so it was able to extract the "likely" build failure reason, we could even present it in the web UI.
We have recently got an RFE in #packit for RPM builds in Copr. The suggestion was to highlight errors so the user can easily spot what went wrong and correct it. This would greatly improve user experience, especially for folks who don't have a ton of experience with RPM builds.
This could also be a great opportunity to even apply some of the machine learning here (as @xsuchy hinted a few weeks ago).
Related: https://pagure.io/copr/copr/issue/401
Originally reported to us by @ssbarnea