Martin Ågren <martin.agren@xxxxxxxxx> writes: > Martin Ågren (11): > sha1_file: do not leak `lock_file` > treewide: prefer lockfiles on the stack > lockfile: fix documentation on `close_lock_file_gently()` > tempfile: fix documentation on `delete_tempfile()` > cache-tree: simplify locking logic > apply: move lockfile into `apply_state` > apply: remove `newfd` from `struct apply_state` > cache.h: document `write_locked_index()` > read-cache: require flags for `write_locked_index()` > read-cache: don't leave dangling pointer in `do_write_index()` > read-cache: roll back lock on error with `COMMIT_LOCK` > > apply.c | 25 ++++++++----------------- > apply.h | 8 +++----- > builtin/am.c | 27 ++++++++++++--------------- > builtin/apply.c | 4 +--- > builtin/checkout.c | 14 ++++++-------- > builtin/clone.c | 7 +++---- > builtin/diff.c | 7 +++---- > builtin/difftool.c | 1 - > cache-tree.c | 12 ++++-------- > cache.h | 19 +++++++++++++++++++ > config.c | 17 ++++++++--------- > git-compat-util.h | 7 ++++++- > lockfile.h | 4 ++-- > merge-recursive.c | 6 +++--- > merge.c | 8 +++----- > read-cache.c | 26 ++++++++++++++------------ > sequencer.c | 1 - > sha1_file.c | 16 +++++++--------- > tempfile.h | 8 ++++---- > wt-status.c | 8 ++++---- > 20 files changed, 110 insertions(+), 115 deletions(-) That's quite a lot of changes to the low-level code. I'll need to revisit this topic later once again, but from a cursory read did not find anything glaringly wrong in it. Thanks for working on this.
