Thread (193 messages) flat view 193 messages, 3 authors, 2021-03-21
STALE1974d

Revision v2 of 2 in this series.

Revisions (2)
  1. v2 current
  2. v4 [diff vs current]

[PATCH v3 06/32] fast-import tests: test for sorting dir/file foo v.s. foo.txt

From: Ævar Arnfjörð Bjarmason <hidden>
Date: 2021-03-16 15:59:38
Subsystem: the rest · Maintainer: Linus Torvalds

Add a missing test for the sorting of "foo" v.s. "foo.txt" where "foo"
can be either a file or a directory. If it's a file then "foo" should
sort before "foo.txt", and the other way around if "foo" is a
directory.

See [1] for a reply to a patch of mine introducing such a
regression. We now finally have a test for this code added back in
463acbe1c6 (Added tree and commit writing to fast-import.,
2006-08-14).

This tests both the tecmp1() and tecmp0() functions introduced inn
4cabf8583f (Implemented tree delta compression in fast-import.,
2006-08-28).

This will catch cases where the "mode" is the same, or reversed
between a & b in both functions.

There was an existing test for the tecmp1() function(s) just above
this one added here.

That existing test was added in e741130386 (New fast-import test case
for valid tree sorting, 2007-03-12) and would catch cases where
entries were reversed, but not if their mode (or rather, type) was the
the same or otherwise wrong value.

There were no tests at all for the tecmp0() function. As with the
tecmp1() test the new test will catch cases where the "mode" is the
same (e.g. "1"), or if the two are reversed. It won't catch a "return
0" from the function (i.e. already sorted), that case requires
tecmp1() to also be broken.

1. https://lore.kernel.org/git/CABPp-BEZfG_pNNTWsnxEa1xG+kgpQfpXj=KHBkCBAMikrXj-aA@mail.gmail.com/ (local)

Signed-off-by: Ævar Arnfjörð Bjarmason <redacted>
---
 t/t9300-fast-import.sh | 87 ++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 87 insertions(+)
diff --git a/t/t9300-fast-import.sh b/t/t9300-fast-import.sh
index 5c47ac4465..8bafb8cc51 100755
--- a/t/t9300-fast-import.sh
+++ b/t/t9300-fast-import.sh
@@ -955,6 +955,93 @@ test_expect_success 'L: verify internal tree sorting' '
 	test_cmp expect actual
 '
 
+test_expect_success 'L: verify internal tree sorting on bad input (tecmp1)' '
+	test_create_repo L1-0 &&
+
+	cat >input <<-INPUT_END &&
+	blob
+	mark :1
+	data 0
+
+	reset refs/heads/L1-0
+	commit refs/heads/L1-0
+	committer $GIT_COMMITTER_NAME <$GIT_COMMITTER_EMAIL> $GIT_COMMITTER_DATE
+	data <<COMMIT
+	create L1-0
+	COMMIT
+	M 100644 :1 x.txt
+	M 100644 :1 x/y.txt
+	M 100644 :1 z.txt
+	M 100644 :1 z
+	INPUT_END
+
+	cat >expected <<-EXPECT_END &&
+	x.txt
+	x
+	z
+	z.txt
+	EXPECT_END
+
+	git -C L1-0 fast-import <input &&
+	git -C L1-0 ls-tree L1-0 >tmp &&
+	cut -f 2 <tmp >actual &&
+	test_cmp expected actual &&
+	git -C L1-0 fsck 2>err &&
+	# Would happen if tecmp1() were broken
+	! grep "error in tree .*: treeNotSorted: " err
+'
+
+test_expect_success 'L: verify internal tree sorting on bad input (tecmp0)' '
+
+	test_create_repo L1-1 &&
+
+	cat >input <<-INPUT_END &&
+	blob
+	mark :1
+	data <<EOF
+	some data
+	EOF
+
+	reset refs/heads/L1-1
+	commit refs/heads/L1-1
+	committer $GIT_COMMITTER_NAME <$GIT_COMMITTER_EMAIL> $GIT_COMMITTER_DATE
+	data <<COMMIT
+	create L1-1
+	COMMIT
+	M 100644 :1 x.txt
+	M 100644 :1 x/y.txt
+	M 100644 :1 z.txt
+	M 100644 :1 z
+
+	commit refs/heads/L1-1
+	committer $GIT_COMMITTER_NAME <$GIT_COMMITTER_EMAIL> $GIT_COMMITTER_DATE
+	data <<COMMIT
+	update L1-1
+	COMMIT
+	M 100644 :1 another.txt
+	M 100644 :1 x.txt
+	M 100644 :1 x/y.txt
+	M 100644 :1 z.txt
+	M 100644 :1 z
+	INPUT_END
+
+	cat >expected <<-EXPECT_END &&
+	another.txt
+	x.txt
+	x
+	z
+	z.txt
+	EXPECT_END
+
+	git -C L1-1 fast-import <input &&
+	git -C L1-1 ls-tree L1-1 >tmp 2>err &&
+	# Would happen if tecmp0() passed a fixed mode
+	! grep "fatal: not a tree object" err &&
+	cut -f 2 <tmp >actual &&
+	test_cmp expected actual &&
+	git -C L1-1 fsck
+'
+
 test_expect_success 'L: nested tree copy does not corrupt deltas' '
 	cat >input <<-INPUT_END &&
 	blob
-- 
2.31.0.256.gf0ddda3145
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help