On Thu, 19 Feb 2026 17:23:55 GMT, David Beaumont <[email protected]> wrote:
>> Okay, so the intention is that it tests the file system view of the run-time >> image with preview classes. Run it with preview features enabled and >> disabled, and ensure that the .class resources contain the expected class >> bytes. That's fine. >> >> The directory is "jrtfs" so it's clear it's testing jrt-fs, so that will >> help shorten the name. It would be helpful if the test had "PreviewFeatures" >> in the name. I think the test summary for the two runs can just focus on >> testing that the resource contains the class bytes for class compiled with >> or without preview features. This is because there isn't any notion of >> "default runtime JRT file system" and the test only runs on the JDK under >> test (the test has no control over this). >> >> Note that testParity (maybe rename to testClassBytes?) shouldn't need to >> translate IOException. >> >> In passing, the make files run the tests with -esa -ea by default so we >> usually don't need to include these in the test description > > Oh blimey, how did I not think to add `-esa -ea`. Thank you for spotting. > I did rename methods in the last push and update the test description. I'll > see if further rewording is worth it based on this comment thread. My use of "default runtime" was to distinguish it from opening other jimage files via URI, which wouldn't be expected match the runtime classes. Instead of "JrtFileSystemClassParityTest" how about "ClassResourcesParityTest". Having preview in the name is a bit odd to me, since in one invocation it's not using preview mode. Also, it does need to translate the IOException because it's called as a lambda via a method reference. ------------- PR Review Comment: https://git.openjdk.org/valhalla/pull/2032#discussion_r2829361396
