On Thu, 15 Aug 2024 15:44:18 GMT, Aleksey Shipilev <sh...@openjdk.org> wrote:
> Honestly, I am ambivalent about this. This only captures the Hotspot build > errors, which I think are rare for Alpine, judging by looking around the JBS. > And it makes sense: the build failures looks basically around compiling > against weird libc APIs that might not be supported by musl. I think most of > the issues with Alpine are runtime problems in tests? > > So it looks that building only Hotspot is in weird spot: we spend time > without getting much of the testing off it. How bad would it be to run > testing with Alpine? That would probably not be very well received, though, > if no one could debug Alpine test failures... > > Tangentially, I think we should start thinking about extended workflows that > are _not_ run by default PR-triggers, so you could potentially run them if > you wanted for Alpine-specific fixes. This would making the decisions like > this rather easy. I was actually going to try building the whole JDK and see what the difference in build time was. If it's not significant then it might be a good first step to just add full build testing. We could always follow up with test running at a later stage if people feel it would be important? ------------- PR Comment: https://git.openjdk.org/jdk/pull/20577#issuecomment-2291582943
