Re: [GSoC][PATCH v3 2/3] t0000: avoid using pipes
From: jonathan chang <hidden>
Date: 2019-03-24 11:26:47
On Mon, Mar 18, 2019 at 12:47 AM Ævar Arnfjörð Bjarmason [off-list ref] wrote:
On Sun, Mar 17 2019, Jonathan Chang wrote:quoted
The exit code of the upstream in a pipe is ignored thus we should avoid using it. By writing out the output of the git command to a file, we can test the exit codes of both the commands. Signed-off-by: Jonathan Chang <redacted> --- t/t0000-basic.sh | 28 ++++++++++++++-------------- 1 file changed, 14 insertions(+), 14 deletions(-)diff --git a/t/t0000-basic.sh b/t/t0000-basic.sh index 53821f5817..47666b013e 100755 --- a/t/t0000-basic.sh +++ b/t/t0000-basic.sh@@ -1118,27 +1118,25 @@ P=$(test_oid root) test_expect_success 'git commit-tree records the correct tree in a commit' ' commit0=$(echo NO | git commit-tree $P) && - tree=$(git show --pretty=raw $commit0 | - sed -n -e "s/^tree //p" -e "/^author /q") && + git show --pretty=raw $commit0 >actual && + tree=$(sed -n -e "s/^tree //p" -e "/^author /q" actual) && test "z$tree" = "z$P"This change is an improvement just changing the "git" invocations. But I wonder as we're reviewing this / churning this if we couldn't also modernize this style to just: git .. >tmp && sed -n -e <tmp >actual && test_must_be_empty actual
Do you mean something like this:
- git show --pretty=raw $commit0 >actual &&
- tree=$(sed -n -e "s/^tree //p" -e "/^author /q" actual) &&
- test "z$tree" = "z$P"
+ git show --pretty=raw $commit0 >tmp &&
+ sed -n -e "/$P/d" -e "s/^tree //p" -e "/^author /q" tmp >actual &&
+ test_must_be_empty actual
It works. But the semantic is different if we use test_must_be_empty.
I wonder if you mean test_cmp because I found some commits[1 2 3]
that changes 'test "z...' to use test_cmp with 'git log -G 'test "z' --oneline'
and git-show. However, they are all around 2013, so I'm not so sure
this is what you mean either.
I did found some use of sed's 'd' function in conjunction with
test_must_be_empty, using:
git grep -A 5 'sed .*/d' | grep -B 5 'test_must_be_empty'
However, they don't use parameter expansion in sed.
There are some places where parameter expansion is used in sed, but
that would require test_cmp in this case, and would need to write to
another file to compare.
Maybe this 'test "z$A" = "z$B"' syntax is fine, yet most of the existing
usages are added around 12 years ago according git-blame I saw on
github.
[1]: 03c893cbf9 ("t1006: modernize output comparisons", 2013-07-10)
[2]: 848575d833 ("push test: simplify check of push result", 2013-03-18)
[3]: ed838e6615 ("t1300: style updates", 2012-10-23)
quoted
@@ -1162,12 +1161,13 @@ test_expect_success 'very long name in the index handled sanely' ' >path4 && git update-index --add path4 && ( - git ls-files -s path4 | - sed -e "s/ .*/ /" | + git ls-files -s path4 >actual && + sed -e "s/ .*/ /" actual | tr -d "\012" && echo "$a" ) | git update-index --index-info && - len=$(git ls-files "a*" | wc -c) && + git ls-files "a*" >actual && + len=$(wc -c <actual) && test $len = 4098Ditto. Maybe the initial author wanted to avoid writing out 4k lines, but now that we're doing so anyway...
This is 'wc -c', so I think I don't have to modify it? By the way, I found 2 lines that can be changed to use test_must_be_empty in t0000-basic.sh. I paste the diff below: ---
@@ -51,7 +51,7 @@ test_expect_success 'verify that the running shellsupports "local"' '
test_expect_success '.git/objects should be empty after git init in
an empty repo' '
find .git/objects -type f -print >should-be-empty &&
- test_line_count = 0 should-be-empty
+ test_must_be_empty should-be-empty
'
# also it should have 2 subdirectories; no fan-out anymore, pack, and info.@@ -1110,7 +1110,7 @@ test_expect_success 'git update-index --refreshshould succeed' '
test_expect_success 'no diff after checkout and git update-index --refresh' '
git diff-files >current &&
- cmp -s current /dev/null
+ test_must_be_empty current
'