On 8/23/2021 4:10 AM, Johannes Berg wrote: > On Mon, 2021-08-23 at 01:05 -0700, Junio C Hamano wrote: >> Johannes Berg <johannes@xxxxxxxxxxxxxxxx> writes: >> >>> I just needed to go outside the current test git directory, the tests >>> are running in a way that the current working directory is already the >>> git tree I'm operating in. >>> >>>> Even if you mean to use "cd", >>>> please do so within a sub-shell. >>> >>> I thought about it, but clearly all the tests are run in a sub-shell, so >>> it didn't seem necessary? But happy to change, I don't really care >>> either way. >> >> Please learn to care before you write your next test, then ;-) > > Hey now, I'm fixing your segfaults ;-) > >> These tests are not run in a sub-shell; they are eval'ed, so that >> the assignment they make to variables can persist and affect the >> next test piece. > > Makes sense. FWIW, the test *did* restore the CWD so things worked, This assumes that your test completes to run the second "cd". > and > subshells are actually ugly (need to import test-lib-functions.sh again > if you want to use those), but I'll make it work somehow. We just add subshells this way: test_expect_success 'test name' ' prep_step && ( # now in a subshell cd wherever && do things # don't need to cd again ) && continue test ' > More importantly, how do you feel about the "cd /"? > > The tests are always run in a place where there's a parent git folder > (even if it's git itself), so you cannot reproduce the segfault in a > test without the "cd /", though I guess "cd /tmp" would also work or > something, but "cd /" felt pretty safe, hopefully not many people have > "/.git" on their system. Don't leave the directory your test is set up to run in. Git has a very large test suite full of examples to use for inspiration. If you do not see a pattern used within the test suite, then there is probably good reason to avoid that pattern. Thanks, -Stolee
