Follow-up a preceding commit (pack-write.c: rename `.idx` file into place last, 2021-08-16)[1] and rename the *.idx file in-place after we write the *.bitmap. The preceding commit fixed the issue of *.idx being written before *.rev files, but did not do so for *.idx files. See 7cc8f971085 (pack-objects: implement bitmap writing, 2013-12-21) for commentary at the time when *.bitmap was implemented about how those files are written out, nothing in that commit contradicts what's being done here. Note that the referenced earlier commit[1] is overly optimistic about "clos[ing the] race", i.e. yes we'll now write the files in the right order, but we might still race due to our sloppy use of fsync(). See the thread at [2] for a rabbit hole of various discussions about filesystem races in the face of doing and not doing fsync() (and if doing fsync(), not doing it properly). 1. https://lore.kernel.org/git/a6a4d2154e83d41c10986c5f455279ab249a910c.1630461918.git.me@xxxxxxxxxxxx/ 2. https://lore.kernel.org/git/8735qgkvv1.fsf@xxxxxxxxxxxxxxxxxxx/ Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@xxxxxxxxx> --- builtin/pack-objects.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/builtin/pack-objects.c b/builtin/pack-objects.c index 33567aaa74..396240c25a 100644 --- a/builtin/pack-objects.c +++ b/builtin/pack-objects.c @@ -1249,7 +1249,6 @@ static void write_pack_file(void) stage_tmp_packfiles(&tmpname, pack_tmp_name, written_list, nr_written, &pack_idx_opts, hash, &idx_tmp_name); - rename_tmp_packfile_idx(&tmpname, hash, &idx_tmp_name); if (write_bitmap_index) { struct strbuf sb = STRBUF_INIT; @@ -1268,6 +1267,8 @@ static void write_pack_file(void) strbuf_release(&sb); } + rename_tmp_packfile_idx(&tmpname, hash, &idx_tmp_name); + free(idx_tmp_name); strbuf_release(&tmpname); free(pack_tmp_name); -- 2.33.0.819.gea1b153a43c
