Re: [PATCH v2 2/3] sequencer: add a "goto cleanup" to do_reset()
- From
Elijah Newren <newren@gmail.com>
- Date
- Oct 7, 2021, 16:06 UTC
- Message-ID
- <CABPp-BH=3DP-dXRCphY53-3eZd1TU8h5GY_M12nnbEGm-UYB9Q@mail.gmail.com>
- In-Reply-To
- <patch-v2-2.3-1d5f5e9fff0-20211007T094019Z-avarab@gmail.com>
On Thu, Oct 7, 2021 at 2:46 AM Ævar Arnfjörð Bjarmason <avarab@gmail.com> wrote:
Show 5 quoted lines
> > Restructure code that's mostly added in 9055e401dd6 (sequencer: > introduce new commands to reset the revision, 2018-04-25) to avoid > code duplication, and to make freeing other resources easier in a > subsequent commit.
Since the first goto you add precedes the call to setup_unpack_trees_porcelain(), the next commit will introduce a case where some code calls clear_unpack_trees_porcelain() without having first called setup_unpack_trees_porcelain(). With the current code that is fine since you do initialize unpack_trees_opts to { 0 } first, and clear_unpack_trees_porcelain() doesn't rely on any special setup, but I wonder if the naming similarity of the two functions might lead to future authors to presume they are always called in pairs.
I wonder if we should rename clear_unpack_trees_porcelain() to e.g. unpack_trees_clear() just to avoid this problem, and then include the other bits of your previous round to just make all callers of unpack_trees() call unpack_trees_clear() when they are done.
Or, maybe we could just make unpack_trees() call unpack_trees_clear() automatically and remove that responsibility from callers...except that merge-recursive is weird and special and would need a new function (unpack_trees_without_automatic_cleanup() or something like that?) since it's the one caller that needs to use unpack_trees_options data after calling unpack_trees().
So, multiple competing ideas here, and all for theoretical future code safety. As such, if any of it sounds like too much of a pain, I'm fine with punting on it.
Show 6 quoted lines
> It's safe to initialize "tree_desc" to be zero'd out in order to > unconditionally free desc.buffer, it won't be initialized on the first > couple of "goto"'s. > > There are three earlier "return"'s in this function that I'm not > bothering to covert,
s/covert/convert/
> those don't need to rollback anything,
They aren't currently rolling back the lockfile right now, but I think they should. Those error cases are pretty unlikely to happen (hard disk full?), but if they did, they'd leave the lock file in place.
Granted, that's a pre-existing problem rather than a problem introduced by your patch, so it doesn't need to be part of this series, but the commit message should at least be tweaked.
> or free any resources
True.
> , so let's leave, even though they could safely "goto > cleanup" as well.
Show 84 quoted lines
>
> Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
> ---
> sequencer.c | 32 +++++++++++++-------------------
> 1 file changed, 13 insertions(+), 19 deletions(-)
>
> diff --git a/sequencer.c b/sequencer.c
> index 6872b7b00a4..457eba4ab10 100644
> --- a/sequencer.c
> +++ b/sequencer.c
> @@ -3650,7 +3650,7 @@ static int do_reset(struct repository *r,
> struct strbuf ref_name = STRBUF_INIT;
> struct object_id oid;
> struct lock_file lock = LOCK_INIT;
> - struct tree_desc desc;
> + struct tree_desc desc = { 0 };
> struct tree *tree;
> struct unpack_trees_options unpack_tree_opts;
> int ret = 0;
> @@ -3684,10 +3684,8 @@ static int do_reset(struct repository *r,
> strbuf_addf(&ref_name, "refs/rewritten/%.*s", len, name);
> if (get_oid(ref_name.buf, &oid) &&
> get_oid(ref_name.buf + strlen("refs/rewritten/"), &oid)) {
> - error(_("could not read '%s'"), ref_name.buf);
> - rollback_lock_file(&lock);
> - strbuf_release(&ref_name);
> - return -1;
> + ret = error(_("could not read '%s'"), ref_name.buf);
> + goto cleanup;
> }
> }
>
> @@ -3703,24 +3701,18 @@ static int do_reset(struct repository *r,
> init_checkout_metadata(&unpack_tree_opts.meta, name, &oid, NULL);
>
> if (repo_read_index_unmerged(r)) {
> - rollback_lock_file(&lock);
> - strbuf_release(&ref_name);
> - return error_resolve_conflict(_(action_name(opts)));
> + ret = error_resolve_conflict(_(action_name(opts)));
> + goto cleanup;
> }
>
> if (!fill_tree_descriptor(r, &desc, &oid)) {
> - error(_("failed to find tree of %s"), oid_to_hex(&oid));
> - rollback_lock_file(&lock);
> - free((void *)desc.buffer);
> - strbuf_release(&ref_name);
> - return -1;
> + ret = error(_("failed to find tree of %s"), oid_to_hex(&oid));
> + goto cleanup;
> }
>
> if (unpack_trees(1, &desc, &unpack_tree_opts)) {
> - rollback_lock_file(&lock);
> - free((void *)desc.buffer);
> - strbuf_release(&ref_name);
> - return -1;
> + ret = -1;
> + goto cleanup;
> }
>
> tree = parse_tree_indirect(&oid);
> @@ -3728,13 +3720,15 @@ static int do_reset(struct repository *r,
>
> if (write_locked_index(r->index, &lock, COMMIT_LOCK) < 0)
> ret = error(_("could not write index"));
> - free((void *)desc.buffer);
>
> if (!ret)
> ret = update_ref(reflog_message(opts, "reset", "'%.*s'",
> len, name), "HEAD", &oid,
> NULL, 0, UPDATE_REFS_MSG_ON_ERR);
> -
> +cleanup:
> + free((void *)desc.buffer);
> + if (ret < 0)
> + rollback_lock_file(&lock);
> strbuf_release(&ref_name);
> return ret;
> }
> --
> 2.33.0.1446.g6af949f83bd
>