| Message ID | 4eab751a3cf00bbffaf4b1084928dad264fa1572.1577454401.git.liu.denton@gmail.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | t: replace incorrect test_must_fail usage (part 2) | expand |
On Fri, Dec 27, 2019 at 8:47 AM Denton Liu <liu.denton@gmail.com> wrote: > Before, we were running `test_must_fail do_checkout`. However, > `test_must_fail` should only be used on git commands. Teach > do_checkout() to accept `!` as a potential first argument which will > prepend `test_must_fail` to the enclosed git command and skips the > remainder of the function. There's a grammatical problem here. s/skips/skip/ is one way to fix it. Use imperative mood when writing commit messages. Drop words such as "before" and "were". For instance: Stop using test_must_fail() with non-Git commands because... (Same comment applies to pretty much all commit messages in this series.) > This increases the granularity of the test as, instead of blindly > checking that do_checkout() failed, we check that only the specific > expected invocation of git fails. This may be a case of trying to describe in prose too much of what is better described by the code itself. As a reviewer, I spent more time trying to figure out what this was saying that I did merely looking at the code and comprehending why the two checks following the git-checkout invocation should be skipped. Consequently, I lean toward dropping "...and skips the remainder..." through the end of the commit message. More below... > Signed-off-by: Denton Liu <liu.denton@gmail.com> > --- > diff --git a/t/t2018-checkout-branch.sh b/t/t2018-checkout-branch.sh > @@ -13,6 +13,12 @@ test_description='checkout' > # > # If <checkout options> is not specified, "git checkout" is run with -b. > do_checkout () { > + should_fail= && > + if test "x$1" = "x!" > + then > + should_fail=test_must_fail && > + shift > + fi && You forgot to update the function comment to talk about the new optional "!" argument. > @@ -26,10 +32,13 @@ do_checkout () { > - git checkout $opts $exp_branch $exp_sha && > + $should_fail git checkout $opts $exp_branch $exp_sha && If I read this literally, it says that the git checkout should always fail. A more wisely chosen variable name would help to alleviate this problem. When you start parameterizing the actual invocation of a command like this (I'm not talking about the command arguments which are also parameterized), the abstraction level and cognitive load increase... > - test $exp_ref = $(git rev-parse --symbolic-full-name HEAD) && > - test $exp_sha = $(git rev-parse --verify HEAD) > + if test -z "$should_fail" > + then > + test $exp_ref = $(git rev-parse --symbolic-full-name HEAD) && > + test $exp_sha = $(git rev-parse --verify HEAD) > + fi > } You could reduce the cognitive load by making the code easier to understand at-a-glance (though at the cost of a minor bit of duplication) by structuring it instead like this: if test -n "$should_fail" then test_must_fail git checkout $opts $exp_branch $exp_sha else git checkout $opts $exp_branch $exp_sha && test $exp_ref = $(git rev-parse --symbolic-full-name HEAD) && test $exp_sha = $(git rev-parse --verify HEAD) fi where 'should_fail' is either empty or non-empty depending upon whether "!" was supplied as an argument. (And, when coded this way, "should_fail" is a reasonable variable name.)
diff --git a/t/t2018-checkout-branch.sh b/t/t2018-checkout-branch.sh index e6b852939c..43551332ed 100755 --- a/t/t2018-checkout-branch.sh +++ b/t/t2018-checkout-branch.sh @@ -13,6 +13,12 @@ test_description='checkout' # # If <checkout options> is not specified, "git checkout" is run with -b. do_checkout () { + should_fail= && + if test "x$1" = "x!" + then + should_fail=test_must_fail && + shift + fi && exp_branch=$1 && exp_ref="refs/heads/$exp_branch" && @@ -26,10 +32,13 @@ do_checkout () { opts="$3" fi - git checkout $opts $exp_branch $exp_sha && + $should_fail git checkout $opts $exp_branch $exp_sha && - test $exp_ref = $(git rev-parse --symbolic-full-name HEAD) && - test $exp_sha = $(git rev-parse --verify HEAD) + if test -z "$should_fail" + then + test $exp_ref = $(git rev-parse --symbolic-full-name HEAD) && + test $exp_sha = $(git rev-parse --verify HEAD) + fi } test_dirty_unmergeable () { @@ -92,7 +101,7 @@ test_expect_success 'checkout -b to a new branch, set to an explicit ref' ' test_expect_success 'checkout -b to a new branch with unmergeable changes fails' ' setup_dirty_unmergeable && - test_must_fail do_checkout branch2 $HEAD1 && + do_checkout ! branch2 $HEAD1 && test_dirty_unmergeable ' @@ -126,7 +135,7 @@ test_expect_success 'checkout -f -b to a new branch with mergeable changes disca test_expect_success 'checkout -b to an existing branch fails' ' test_when_finished git reset --hard HEAD && - test_must_fail do_checkout branch2 $HEAD2 + do_checkout ! branch2 $HEAD2 ' test_expect_success 'checkout -b to @{-1} fails with the right branch name' ' @@ -165,7 +174,7 @@ test_expect_success 'checkout -B to an existing branch with unmergeable changes git checkout branch1 && setup_dirty_unmergeable && - test_must_fail do_checkout branch2 $HEAD1 -B && + do_checkout ! branch2 $HEAD1 -B && test_dirty_unmergeable '
Before, we were running `test_must_fail do_checkout`. However, `test_must_fail` should only be used on git commands. Teach do_checkout() to accept `!` as a potential first argument which will prepend `test_must_fail` to the enclosed git command and skips the remainder of the function. This increases the granularity of the test as, instead of blindly checking that do_checkout() failed, we check that only the specific expected invocation of git fails. Signed-off-by: Denton Liu <liu.denton@gmail.com> --- t/t2018-checkout-branch.sh | 21 +++++++++++++++------ 1 file changed, 15 insertions(+), 6 deletions(-)