git/list[1] front-page[2] threads[3] people[4] search[5] about
 

Strange diff-index with fsmonitor, submodules

From
JTJonathan Tan <jonathantanmy@google.com>
Date
Aug 29, 2023, 00:56 UTC
Message-ID
<20230829005606.136615-1-jonathantanmy@google.com>

There is a strange interaction where diff-index not only produces different results when run with and without fsmonitor, but produces different results for 2 entries that as far as I can tell, should behave identically (sibling files in the same directory - file_11 and file_12, and both of these filenames are only mentioned once each in the entire test).

You can see this with this patch:
diff --git a/t/t7527-builtin-fsmonitor.sh b/t/t7527-builtin-fsmonitor.sh
index 0c241d6c14..e9e5e32016 100755
--- a/t/t7527-builtin-fsmonitor.sh
+++ b/t/t7527-builtin-fsmonitor.sh
@@ -809,6 +809,11 @@ my_match_and_clean () {
                status --porcelain=v2 >actual.without &&
        test_cmp actual.with actual.without &&
 
+       git -C super --no-optional-locks diff-index --name-status HEAD >actual.with &&
+       git -C super --no-optional-locks -c core.fsmonitor=false \
+               diff-index --name-status HEAD >actual.without &&
+       test_cmp actual.with actual.without &&
+
        git -C super/dir_1/dir_2/sub reset --hard &&
        git -C super/dir_1/dir_2/sub clean -d -f
 }
@@ -837,6 +842,7 @@ test_expect_success 'submodule always visited' '
        # some dirt in the submodule and confirm matching output.
 
        # Completely clean status.
+       echo Now running for clean status &&
        my_match_and_clean &&

and this is the output:

++ echo Now running for clean status
Now running for clean status
++ my_match_and_clean
++ git -C super --no-optional-locks status --porcelain=v2
++ git -C super --no-optional-locks -c core.fsmonitor=false status --porcelain=v2
++ test_cmp actual.with actual.without
++ test 2 -ne 2
++ eval 'diff -u' '"$@"'
+++ diff -u actual.with actual.without
++ git -C super --no-optional-locks diff-index --name-status HEAD
++ git -C super --no-optional-locks -c core.fsmonitor=false diff-index --name-status HEAD
++ test_cmp actual.with actual.without
++ test 2 -ne 2
++ eval 'diff -u' '"$@"'
+++ diff -u actual.with actual.without
--- actual.with	2023-08-29 00:39:26
+++ actual.without	2023-08-29 00:39:26
@@ -1 +0,0 @@
-D	dir_1/file_11
error: last command exited with $?=1
not ok 61 - submodule always visited

Notice that with fsmonitor, diff-index reports a "D" line that is not
present when fsmonitor is off. To add to that, it only reports "D" for
file_11 when I would expect that if it reported file_11, it would report
file_12 as well.

I'll continue investigating this myself, but does anyone know what is
going on?
Next: Junio C Hamano
Message 1 of 18 in “Strange diff-index with fsmonitor, submodules”
  1. Jonathan TanAug 29, 2023
  2. Junio C HamanoAug 29, 2023
  3. Jeff HostetlerAug 29, 2023
  4. Jeff HostetlerAug 29, 2023
  5. Jonathan TanAug 29, 2023
  6. [diff-lib] Fix check_removed when fsmonitor is onJosip Sokcevic, Sep 6, 2023
  7. Jonathan TanSep 6, 2023
  8. diff-lib: Fix check_removed when fsmonitor is onJosip Sokcevic, Sep 7, 2023
  9. Jonathan TanSep 7, 2023
  10. Junio C HamanoSep 7, 2023
  11. Josip SokcevicSep 7, 2023
  12. Junio C HamanoSep 8, 2023
  13. diff-lib: Fix check_removed when fsmonitor is onJosip Sokcevic, Sep 11, 2023
  14. Junio C HamanoSep 11, 2023
  15. Josip SokcevicSep 12, 2023
  16. Junio C HamanoSep 12, 2023
  17. Josip SokcevicSep 14, 2023
  18. Junio C HamanoSep 18, 2023

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.