On Mon, 2 Sep 2024 07:20:50 GMT, Johan Vos <j...@openjdk.org> wrote: > When building webkit, the perl script in > `modules/javafx.web/src/main/native/Tools/Scripts/build-webkit` will check > for a `-j` argument in `$makeArgs`. If that doesn't exist, it will default to > `numberOfCPUs()`. > This PR passes the `NUM_COMPILE_THREADS` from `build.gradle` to the > `build-webkit` script. When not setting `NUM_COMPILE_THREADS` explicitly,it > will default to the number of CPU's as well, hence it will have the same > behavior as without this PR. > When setting `NUM_COMPILE_THREADS` explicitly, the provided number will be > used by cmake as well now.
As you say, option 3 is the cleanest, however I believe it will very quickly morph into option 4. I am reasonably certain that [JDK-8089417](https://bugs.openjdk.org/browse/JDK-8089417) can be closed as "Cannot reproduce" because: * The font T2K code that was hitting the failure is no longer part of our repo; T2K was a closed-source font rasterizer that we removed in JDK 10 via [JDK-8187147](https://bugs.openjdk.org/browse/JDK-8187147). * Xcode switched from gcc to clang by default several years ago I will locally remove the workaround, and then run the following in a loop a few hundred times to check whether we see any issues: $ gradle cleanNative sdk I did it a few times on my MacBook M1 and the only impact I see of removing the workaround is that a rebuild of the SDK without media or WebKit (so just rebuilding the graphics / glass native code) goes from 27 seconds to 12 seconds. :) ------------- PR Comment: https://git.openjdk.org/jfx/pull/1554#issuecomment-2327034232
