On Thu, 12 Jan 2023 12:26:16 GMT, Magnus Ihse Bursie <i...@openjdk.org> wrote:
> Hi, I'm back today from vacation and is trying to catch up, so your timing > was excellent. ;) Haha, Merry Christmas and a Happy New Year, hope you had a good time with your family :) > This looks good. I have seen your attempt at fighting with GHA in #11020, and > I am impressed with your patience and stubbornness. :-) I'm glad it panned > out in a solution. I don't think I've ever been able to sleep whenever I've run into a problem with anything about the JDK that doesn't initially work :P Maybe I should apply for a job at Oracle and work full time on the Java Platform, wonder how that will turn out, heh > I deliberately designed the GHA code so that each action that installs some > part returns the path. This means it should be trivial to move where gtest, > jtreg etc is installed as well. I did not specifically study the different > storage roots provided by GHA, so it might very well be the case that all of > these should be stored in the tool cache as well. If this is your conclusion, > feel free to open a new PR where you update them in the same way. It depends, if we are using a GitHub hosted runner (which I am very confident we are) the tool cache would be more appropriate for dumping all our required utils than the source directory > Actually, hold your horses. I checked the GHA run, and there were a lot of > failures on windows... `tar` did not work, and seems it tried to pick up one > from the Git unix env. Uh oh, that isn't good. Part of the reason for the changes here is to fix that issue (I've checked the Actions and this issue is pretty much always there on Windows). Strange, I checked the early GHA runs of this PR and it did fix the problem, why the regression now? Sigh... ------------- PR: https://git.openjdk.org/jdk/pull/11782
