When loosening a pack, the current pack_id gets reused when checkpointing and the import does not terminate. This causes problems after checkpointing as the object table, branch, and tag lists still contains pre-checkpoint references to the recycled pack_id. Merely clearing the object_table as suggested by Jeff King in http://mid.gmane.org/20160517121330.GA7346@xxxxxxxxxxxxxxxxxxxxx is insufficient as the marks set still contains references to object entries. Wrong pack_id references branch and tags lists do not cause errors, but can lead to misleading crash reports and core dumps, so they are also invalidated. Signed-off-by: Eric Wong <e@xxxxxxxxx> --- I started writing a standalone test case; but testing with original failing cases would be greatly appreciated. Still learning my way around the fast-import code... Thanks. fast-import.c | 31 +++++++++++++++++++- t/t9302-fast-import-unpack-limit.sh | 57 +++++++++++++++++++++++++++++++++++++ 2 files changed, 87 insertions(+), 1 deletion(-) diff --git a/fast-import.c b/fast-import.c index 0e8bc6a..b9db4b6 100644 --- a/fast-import.c +++ b/fast-import.c @@ -597,6 +597,33 @@ static struct object_entry *insert_object(unsigned char *sha1) return e; } +static void invalidate_pack_id(unsigned int id) +{ + unsigned int h; + unsigned long lu; + struct tag *t; + + for (h = 0; h < ARRAY_SIZE(object_table); h++) { + struct object_entry *e; + + for (e = object_table[h]; e; e = e->next) + if (e->pack_id == id) + e->pack_id = MAX_PACK_ID; + } + + for (lu = 0; lu < branch_table_sz; lu++) { + struct branch *b; + + for (b = branch_table[lu]; b; b = b->table_next_branch) + if (b->pack_id == id) + b->pack_id = MAX_PACK_ID; + } + + for (t = first_tag; t; t = t->next_tag) + if (t->pack_id == id) + t->pack_id = MAX_PACK_ID; +} + static unsigned int hc_str(const char *s, size_t len) { unsigned int r = 0; @@ -993,8 +1020,10 @@ static void end_packfile(void) cur_pack_sha1, pack_size); if (object_count <= unpack_limit) { - if (!loosen_small_pack(pack_data)) + if (!loosen_small_pack(pack_data)) { + invalidate_pack_id(pack_id); goto discard_pack; + } } close(pack_data->pack_fd); diff --git a/t/t9302-fast-import-unpack-limit.sh b/t/t9302-fast-import-unpack-limit.sh index 0f686d2..a04de14 100755 --- a/t/t9302-fast-import-unpack-limit.sh +++ b/t/t9302-fast-import-unpack-limit.sh @@ -45,4 +45,61 @@ test_expect_success 'bigger packs are preserved' ' test $(find .git/objects/pack -type f | wc -l) -eq 2 ' +test_expect_success 'lookups after checkpoint works' ' + hello_id=$(echo hello | git hash-object --stdin -t blob) && + id="$GIT_COMMITTER_NAME <$GIT_COMMITTER_EMAIL> $GIT_COMMITTER_DATE" && + before=$(git rev-parse refs/heads/master^0) && + ( + cat <<-INPUT_END && + blob + mark :1 + data 6 + hello + + commit refs/heads/master + mark :2 + committer $id + data <<COMMIT + checkpoint after this + COMMIT + from refs/heads/master^0 + M 100644 :1 hello + + # pre-checkpoint + cat-blob :1 + cat-blob $hello_id + checkpoint + # post-checkpoint + cat-blob :1 + cat-blob $hello_id + INPUT_END + + n=0 && + from=$before && + while test x"$from" = x"$before" + do + if test $n -gt 30 + then + echo >&2 "checkpoint did not update branch" + exit 1 + else + n=$(($n + 1)) + fi && + sleep 1 && + from=$(git rev-parse refs/heads/master^0) + done && + cat <<-INPUT_END && + commit refs/heads/master + committer $id + data <<COMMIT + make sure from "unpacked sha1 reference" works, too + COMMIT + from $from + INPUT_END + echo done + ) | git -c fastimport.unpackLimit=100 fast-import --done && + test $(find .git/objects/?? -type f | wc -l) -eq 6 && + test $(find .git/objects/pack -type f | wc -l) -eq 2 +' + test_done -- To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html