On Fri, Sep 08, 2017 at 03:51:42PM +0200, Michael Haggerty wrote: > This is v2 of a patch series to implement reference transactions for > the packed refs-store. Thanks to Stefan, Brandon, Junio, and Peff for > your review of v1 [1]. I believe I have addressed all of your > comments. > > Changes since v1: > > * Patch [01/11]: justify the change better in the log message. Add a > comment explaining why `get_packed_ref_cache()` is being called but > the return value discarded. > > * Patch [05/11]: Lock the `packed-refs` file *after* successfully > creating the (empty) transaction object. This prevents leaving the > file locked if `ref_store_transaction_begin()` fails. > > * Patch [06/11]: New patch, fixing a leak of the `refs_to_prune` > linked list. These all look good to me, including the new patch. > * Patch [07/11]: Reimplement test "no bogus intermediate values during > delete" to work without polling. Also incorporate Junio's change > `s/grep/test_i18ngrep/`. I picked a few nits in the test script. Nothing too serious, but a few things that might be worth addressing. -Peff
