On Tue, Aug 6, 2024 at 12:35 PM Junio C Hamano <gitster@xxxxxxxxx> wrote: > Eric Sunshine <sunshine@xxxxxxxxxxxxxx> writes: > > As a practical example, consider a test with a body like this: > > > > echo nothing >nothing && > > git add nothing && > > git commit -m nothing && > > hash=$(git rev-parse HEAD) && > > ... > > > > If this body is inside a double-quoted string, then `$(git rev-parse > > HEAD)` will be evaluated and its value assigned to `hash` _before_ > > test_expect_success() is called, > > I know it is just your finger slipping, but the variable "hash" is > not assigned to before test_expect_success is called even with the > body inside dq. Ugh, you're quite correct, of course. The "and its value assigned to `hash`" part was one of those last-moment-edits which come back to haunt the author. That's not so much a finger slip as an outright bungle. Thanks for catching and correcting this. > What happens is that the value of HEAD is expanded in the string > that will be evaled by test_expect_success so the 4th line in the > above becomes "hash=3469a23659d8197190d2765cf9f31dec5ab602fa &&"; > as the resulting string is then eval'ed by test_expect_success, > the end result is as you descirbed, i.e., ... > ... $hash does not get the name of the commit object resulting from > the "git commit" command before it. Including an actual example of the resulting string, as you did here, does a much better job of illustrating the issue than my bungled last-moment-edit.
