IRC channel logs

<dsmith>How do you run just one of test-suite/tests/*.test ?

<dsmith>I used to know, but now I can't remember..

<dsmith>This seems to do it: meta/guile --no-auto-compile -L test-suite test-suite/tests/foo.test

<tianshu>Hey, guys. I'm glad to be here!

<rlb>dsmith: see the comments at the top of check-guile

<dsmith>Ok, found one problem. In test-suite/test-suite/lib.scm run-test, in the (fail) case, extra args are applied. But there is no match for that in test-suite/test-suite/lib/automake.scm reporter

<dsmith>So you get a backtrace there

<dsmith>This is when you have a failing test that is expected to pass

<dsmith>The second problem, (after adding a `. args` the 'fail case) is that the "make check" still stops there and doesn't report the summary.

<rlb>can you see what's going on (I think "make V=1 check" may show more)?

<dsmith>That didn't show anything extra.

<dsmith>The Makefile is calling the sh script "drvier" which is exec'ing the "check-guile" script

<dsmith>check-guile is exiting with non-zero

<rlb>You probably found it, but here's where the harness was switched over: df04f5357a0c7146d7ec1c4fcd8c11a42feb5e01 And offhand, I'm not even sure what the automake behavior is expected to be there.

<dsmith>And *that* is exec'ing guile-test

<dsmith>tianshu, Welcome!

<dsmith>If I change the (quit global-pass) to (quit 0) in main in test-suite/guile-test, I do get a summary at the end of make check

<dsmith>And now my most wonderful wife just brought home some Indian cuisine....

<rlb>mmmmmm

<rlb>dsmith: I'll have to refresh to be much help, i.e. remember what the expectations are.

<rlb>You were looking at test-suite/test tests, right, not standalone?

<dsmith>Correct

<dsmith>Ok. Background. Everything works great when there are no failing tests. Or rather every test has the expected outcome (pass, fail skip ...)

<dsmith>You get the nice summary at the end. The standalone tests have a separate summary.

<dsmith>But if test expected to pass fails, the make stops there. No summary at the end.

<dsmith>And I assume also if an expected failure passes, but I didn't test that.

<dsmith>rlb, I must say that having parallel tests is quite an improvement!

<rlb>dsmith: right - messed with that because I wanted *faster* :)

<rlb>And I'd guess that maybe it's just when the whole thing crashes and exits with nonzero status, but I'll have to review.

<rlb>i.e. when any given foo.test does

IRC channel logs

2025-09-06.log