On Tue, 22 Sep 2026 18:05:57 GMT, Nir Lisker <[email protected]> wrote:

>> To Nir's point, this isn't an existing build script. There are no build 
>> scripts for manual tests and all of our other tests use gradle. Right now, 
>> only "apps" use ant, and that's because they are legacy. I wouldn't use that 
>> as an argument for expanding the use of ant beyond the apps dir.
>
>> All my builds are clean builds, I had my share of problems stemming from 
>> stale artifacts.
> 
> Yes, because the Gradle build file is incorrect. I filed 
> [JDK-8391666](https://bugs.openjdk.org/browse/JDK-8391666) because the stale 
> artifacts aren't detected correctly, which results in compromised builds. 
> [JDK-8389921](https://bugs.openjdk.org/browse/JDK-8389921) fixed an example 
> of this with native sources. With all of that, it's still rare that I need to 
> clean every build.
> 
>> Do you want to migrate all the build scripts in the project to use gradle?
> 
> Without fixing the main build file it won't help much because these 
> apps/tests are downstream of the broken build. They'll run a bit better than 
> ant. I have https://github.com/openjdk/jfx/pull/2294 sitting in the review 
> queue for a couple of weeks (not complaining), and 
> https://github.com/openjdk/jfx/pull/2305 ready to go in directly after. Then 
> a couple more in the chain that I'm waiting with until the first one moves.

> it also adds significant overhead [0] and a maintenance burden.
> 
> [0] https://mill-build.org/blog/1-java-compile.html

That's because it's an actual build system, and not an artificial benchmark 
that shows anything and nothing. Once you solve all the problems of a complex 
build, you _will_ have a complex build script and the artificial advantage goes 
away. In fact, ant is _much_ slower than Gradle on very complex builds once you 
account for repeated edit-and-build scenarios.

-------------

PR Review Comment: https://git.openjdk.org/jfx/pull/2156#discussion_r4074874946

Reply via email to