[v3] diff-lib: ignore all outsider if --relative asked

Commit Message

Đoàn Trần Công Danh Aug. 22, 2021, 8:49 a.m. UTC

For diff family commands, we can tell them to exclude changes outside
of some directories if --relative is requested.

In diff_unmerge(), NULL will be returned if the requested path is
outside of the interesting directories, thus we'll run into NULL
pointer dereference in run_diff_files when trying to dereference
its return value.

Checking for return value of diff_unmerge before dereferencing
is not sufficient, though. Since, diff engine will try to work on such
pathspec later.

Let's not run diff on those unintesting entries, instead.
As a side effect, by skipping like that, we can save some CPU cycles.

Reported-by: Thomas De Zeeuw <[e-mail]>
Tested-by: Carlo Arenas <[e-mail]>
Signed-off-by: Đoàn Trần Công Danh <[e-mail]>
---

Sorry for the noise, I need to send v3, because:

Change in v3 from v2:
 Fix the conflict marker in failure test. Originally, I wrote the test with:
 "git merge sub1" not "git merge br1"

Range-diff against v2:
1:  0d73c71819 ! 1:  a140b292f9 diff-lib: ignore all outsider if --relative asked
    @@ t/t4045-diff-relative.sh: check_diff_relative_option subdir file2 true --no-rela
     +	 +sub3
     +	++=======
     +	+ sub1
    -+	++>>>>>>> sub1
    ++	++>>>>>>> br1
     +	EOF
     +	git -C subdir diff --relative >out &&
     +	test_cmp expected out

Range-diff of v2 against v1:
1:  57a9edc3af ! 1:  0d73c71819 diff-lib: ignore all outsider if --relative asked
    @@ Commit message
         pointer dereference in run_diff_files when trying to dereference
         its return value.
     
    -    We can simply check for NULL there before dereferencing said
    -    return value.  However, we can do better by not running diff
    -    on those unintesting entries.  Let's do that instead.
    +    Checking for return value of diff_unmerge before dereferencing
    +    is not sufficient, though. Since, diff engine will try to work on such
    +    pathspec later.
    +
    +    Let's not run diff on those unintesting entries, instead.
    +    As a side effect, by skipping like that, we can save some CPU cycles.
     
         Reported-by: Thomas De Zeeuw <[e-mail]>
    +    Tested-by: Carlo Arenas <[e-mail]>
         Signed-off-by: Đoàn Trần Công Danh <[e-mail]>
     
 diff-lib.c               |  4 +++
 t/t4045-diff-relative.sh | 53 ++++++++++++++++++++++++++++++++++++++++
 2 files changed, 57 insertions(+)

Patch

diff --git a/diff-lib.c b/diff-lib.c
index f9eadc4fc1..ca085a03ef 100644
--- a/diff-lib.c
+++ b/diff-lib.c
@@ -117,6 +117,10 @@  int run_diff_files(struct rev_info *revs, unsigned int option)
 		if (!ce_path_match(istate, ce, &revs->prune_data, NULL))
 			continue;
 
+		if (revs->diffopt.prefix &&
+		    strncmp(ce->name, revs->diffopt.prefix, revs->diffopt.prefix_length))
+			continue;
+
 		if (ce_stage(ce)) {
 			struct combine_diff_path *dpath;
 			struct diff_filepair *pair;
diff --git a/t/t4045-diff-relative.sh b/t/t4045-diff-relative.sh
index 61ba5f707f..fab351b48a 100755
--- a/t/t4045-diff-relative.sh
+++ b/t/t4045-diff-relative.sh
@@ -162,4 +162,57 @@  check_diff_relative_option subdir file2 true --no-relative --relative
 check_diff_relative_option . file2 false --no-relative --relative=subdir
 check_diff_relative_option . file2 true --no-relative --relative=subdir
 
+test_expect_success 'setup diff --relative unmerged' '
+	test_commit zero file0 &&
+	test_commit base subdir/file0 &&
+	git switch -c br1 &&
+	test_commit one file0 &&
+	test_commit sub1 subdir/file0 &&
+	git switch -c br2 base &&
+	test_commit two file0 &&
+	git switch -c br3 &&
+	test_commit sub3 subdir/file0
+'
+
+test_expect_success 'diff --relative without change in subdir' '
+	git switch br2 &&
+	test_when_finished "git merge --abort" &&
+	test_must_fail git merge one &&
+	git -C subdir diff --relative >out &&
+	test_must_be_empty out &&
+	git -C subdir diff --relative --name-only >out &&
+	test_must_be_empty out
+'
+
+test_expect_success 'diff --relative --name-only with change in subdir' '
+	git switch br3 &&
+	test_when_finished "git merge --abort" &&
+	test_must_fail git merge sub1 &&
+	test_write_lines file0 file0 >expected &&
+	git -C subdir diff --relative --name-only >out &&
+	test_cmp expected out
+'
+
+test_expect_failure 'diff --relative with change in subdir' '
+	git switch br3 &&
+	br1_blob=$(git rev-parse --short --verify br1:subdir/file0) &&
+	br3_blob=$(git rev-parse --short --verify br3:subdir/file0) &&
+	test_when_finished "git merge --abort" &&
+	test_must_fail git merge br1 &&
+	cat >expected <<-EOF &&
+	diff --cc file0
+	index $br3_blob,$br1_blob..0000000
+	--- a/file0
+	+++ b/file0
+	@@@ -1,1 -1,1 +1,5 @@@
+	++<<<<<<< HEAD
+	 +sub3
+	++=======
+	+ sub1
+	++>>>>>>> br1
+	EOF
+	git -C subdir diff --relative >out &&
+	test_cmp expected out
+'
+
 test_done