On Tue, 7 Oct 2025 at 23:55, Alexander Bantyev <[email protected]> wrote:
> Previously if a test produced a diff, the only way to fix it was to > inspect the test output. > And in some cases, it's non-trivial to figure out which test input corresponds to which expected output. (I'm working on a patch to rationalise the test harness framework, which will make this a *lot* simpler.) We introduce the BASH_TSTOUT_KEEPDIR environment variable which is > expected to point to a directory. > If set, that directory will be populated with output from all failed > tests, with test names as file names. > > This makes debugging or updating the expected test outputs easier. > Whilst I like the idea of keeping the output, this approach seems a bit backwards. Rather, arrange for each test's output to be written to a separate file in a per-run directory, and then if we *don't* want to keep the output, remove the whole directory. I'll work up a patch to do this. -Martin
