| Message ID | 20200812192737.13971-2-shouryashukla.oo@gmail.com (mailing list archive) |
|---|---|
| State | Superseded |
| Headers | show |
| Series | t7401: modernize, cleanup and more | expand |
On Thu, 2020-08-13 at 00:57 +0530, Shourya Shukla wrote: > > [...] > > Using a Git command in the upstream of a pipe might result in us > losing its exit code. So, convert such usages so that they write to > a file and read from them. > While that is a good enough reason to avoid using pipes in places where we look for the exit code of a command like within test_expect_success, I'm not sure if that reason holds for the places that the patch changes. > [...] > > diff --git a/t/t7401-submodule-summary.sh b/t/t7401-submodule- > summary.sh > index 9bc841d085..8ee78bcb69 100755 > --- a/t/t7401-submodule-summary.sh > +++ b/t/t7401-submodule-summary.sh > @@ -16,12 +16,13 @@ add_file () { > owd=$(pwd) > cd "$sm" > for name; do > - echo "$name" > "$name" && > + echo "$name" >"$name" && > git add "$name" && > test_tick && > git commit -m "Add $name" > done >/dev/null > - git rev-parse --verify HEAD | cut -c1-7 > + git rev-parse --verify HEAD >out && > + cut -c1-7 out In any case, I believe we can avoid the 'cut' altogether in both places by doing something like this instead: git rev-parse --short=7 HEAD My quick check shows the test script is happy with this change. > cd "$owd" > } > commit_file () { > @@ -125,7 +126,8 @@ commit_file sm1 && > head3=$( > cd sm1 && > git reset --hard HEAD~2 >/dev/null && > - git rev-parse --verify HEAD | cut -c1-7 > + git rev-parse --verify HEAD >out && > + cut -c1-7 out > ) > > test_expect_success 'modified submodule(backward)' "
Kaartic Sivaraam <kaartic.sivaraam@gmail.com> writes: >> @@ -16,12 +16,13 @@ add_file () { >> owd=$(pwd) >> cd "$sm" >> for name; do >> - echo "$name" > "$name" && >> + echo "$name" >"$name" && >> git add "$name" && >> test_tick && >> git commit -m "Add $name" >> done >/dev/null >> - git rev-parse --verify HEAD | cut -c1-7 >> + git rev-parse --verify HEAD >out && >> + cut -c1-7 out > > In any case, I believe we can avoid the 'cut' altogether in both places > by doing something like this instead: > > git rev-parse --short=7 HEAD Ah, I missed the fact that this was a helper function and most of the error status is discarded anyway. For example, we still run the rev-parse even after the for loop fails. If the focus of this test script were to ensure that rev-parse works correctly, being careful to catch its exit status might have had a good value, but for that, all the other operations that happen in this helper function (including the "what happens when the loop body fails for $name that is not at the end of the argument list?") must also be checked for their exit status in the first place. Since that is not done, and since testing rev-parse should not have to be part of the job for submodule test, the net effect of the above change has quite dubious value---it clobbered a file 'out' that may be used by the caller. Doing "cd" without introducing a subshell is a bit harder to fix, as test_tick relies on the global counter in the topmost process. It can be done, but I do not think it is worth doing here. Most of the users of this helper function call it in var=$(add_file ...) subshell anyway (so test_tick is incrementing the timestamp independently for each caller and discarding the resulting timestamp). As a NEEDSWORK comment added in the series says, this script may need a bit more work. I agree with you that the split of "rev-parse | cut -c1-7" into two statements and clobbering 'out' is a bad change---that part should be reverted. The style change on 'echo "$name" >"$name"' line is OK, though. Thanks. > My quick check shows the test script is happy with this change. > >> cd "$owd" >> } >> commit_file () { >> @@ -125,7 +126,8 @@ commit_file sm1 && >> head3=$( >> cd sm1 && >> git reset --hard HEAD~2 >/dev/null && >> - git rev-parse --verify HEAD | cut -c1-7 >> + git rev-parse --verify HEAD >out && >> + cut -c1-7 out >> ) >> >> test_expect_success 'modified submodule(backward)' "
On 13/08 09:46, Junio C Hamano wrote: > Kaartic Sivaraam <kaartic.sivaraam@gmail.com> writes: > > >> @@ -16,12 +16,13 @@ add_file () { > >> owd=$(pwd) > >> cd "$sm" > >> for name; do > >> - echo "$name" > "$name" && > >> + echo "$name" >"$name" && > >> git add "$name" && > >> test_tick && > >> git commit -m "Add $name" > >> done >/dev/null > >> - git rev-parse --verify HEAD | cut -c1-7 > >> + git rev-parse --verify HEAD >out && > >> + cut -c1-7 out > > > > In any case, I believe we can avoid the 'cut' altogether in both places > > by doing something like this instead: > > > > git rev-parse --short=7 HEAD > > Ah, I missed the fact that this was a helper function and most of > the error status is discarded anyway. For example, we still run the > rev-parse even after the for loop fails. > > If the focus of this test script were to ensure that rev-parse works > correctly, being careful to catch its exit status might have had a > good value, but for that, all the other operations that happen in > this helper function (including the "what happens when the loop body > fails for $name that is not at the end of the argument list?") must > also be checked for their exit status in the first place. > > Since that is not done, and since testing rev-parse should not have > to be part of the job for submodule test, the net effect of the > above change has quite dubious value---it clobbered a file 'out' > that may be used by the caller. > > Doing "cd" without introducing a subshell is a bit harder to fix, as > test_tick relies on the global counter in the topmost process. It > can be done, but I do not think it is worth doing here. Most of the > users of this helper function call it in var=$(add_file ...) > subshell anyway (so test_tick is incrementing the timestamp > independently for each caller and discarding the resulting > timestamp). As a NEEDSWORK comment added in the series says, this > script may need a bit more work. > > I agree with you that the split of "rev-parse | cut -c1-7" into two > statements and clobbering 'out' is a bad change---that part should > be reverted. The style change on 'echo "$name" >"$name"' line is > OK, though. > > Thanks. Understood. I will revert the change. Though, what Kaartic suggested, to do a '--short=7', that will be okay to keep right? Something like: git rev-parse --short=7 HEAD This way we will not need a 'cut'. This change goes as a separate commit obviously.
Shourya Shukla <shouryashukla.oo@gmail.com> writes: > Understood. I will revert the change. Though, what Kaartic suggested, to > do a '--short=7', that will be okay to keep right? Sure, that is a strict improvement to lose an unneeded process, as long as we know HEAD is guaranteed to be unique with 7 hexdigits (otherwise "cut" to strictly 7 and "rev-parse --short=7 HEAD" would produce different results. Thanks.
diff --git a/t/t7401-submodule-summary.sh b/t/t7401-submodule-summary.sh index 9bc841d085..8ee78bcb69 100755 --- a/t/t7401-submodule-summary.sh +++ b/t/t7401-submodule-summary.sh @@ -16,12 +16,13 @@ add_file () { owd=$(pwd) cd "$sm" for name; do - echo "$name" > "$name" && + echo "$name" >"$name" && git add "$name" && test_tick && git commit -m "Add $name" done >/dev/null - git rev-parse --verify HEAD | cut -c1-7 + git rev-parse --verify HEAD >out && + cut -c1-7 out cd "$owd" } commit_file () { @@ -125,7 +126,8 @@ commit_file sm1 && head3=$( cd sm1 && git reset --hard HEAD~2 >/dev/null && - git rev-parse --verify HEAD | cut -c1-7 + git rev-parse --verify HEAD >out && + cut -c1-7 out ) test_expect_success 'modified submodule(backward)' "
The tests in 't7401-submodule-summary.sh' were written a long time ago and has a violation with respect to our CodingGuidelines which is, incorrect spacing in usages of the redirection operator. Using a Git command in the upstream of a pipe might result in us losing its exit code. So, convert such usages so that they write to a file and read from them. Mentored-by: Christian Couder <chriscool@tuxfamily.org> Mentored-by: Kaartic Sivaraam <kaartic.sivaraam@gmail.com> Helped-by: Denton Liu <liu.denton@gmail.com> Helped-by: Taylor Blau <me@ttaylorr.com> Signed-off-by: Shourya Shukla <shouryashukla.oo@gmail.com> --- t/t7401-submodule-summary.sh | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-)