RFR: 7903692: JTREG_HOME and JT_HOME can interfere with jtreg self tests

[Jaikiran Pai]

On Thu, 7 Mar 2024 07:52:51 GMT, Jaikiran Pai <jpai at openjdk.org> wrote:

> Can I please get a review of this change which proposes to fix the issue noted in https://bugs.openjdk.org/browse/CODETOOLS-7903692?
> 
> As noted in that issue, when running the jtreg self tests for the project, if `JT_HOME` (or `JTREG_HOME`) have been configured to point to some specific jtreg installation, then the self tests instead of running against the locally built/changed jtreg will end up running against the jtreg installation pointed to by that environment variable. That effectively means that the changes being tested aren't tested against the actual changed code in jtreg and the test results can be misleading.
> 
> The commit in this PR generates a `make.sh` which resets the `JT_HOME` and `JTREG_HOME` environment variables to be empty so that the `jtreg` script that's launched from the locally built jtreg image will then pick up the correct locally built jtreg image.
> 
> I have run the jtreg self tests locally with this change and they continue to pass. Without this change, some of the tests fail because they end up picking a jtreg installation which doesn't have some recent changes that are available in this git repo (that's what prompted me to investigate this issue).

Hello Christian,

> but could potentially break use-cases on other setups, right?

I'm unaware if running the jtreg self tests was expected to run those tests against anything other than the one being built and tested. If at all we need to support the case where these tests need to be run against a specific jtreg binary, then maybe we could enhance these scripts (and documentation) to expect the user to explicitly set a `JTREG_TEST_BINARY_HOME` (this variable doesn't exist right now), so that the envrionment variable is only known/honoured by the test run?

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

PR Comment: https://git.openjdk.org/jtreg/pull/189#issuecomment-1983083632