From: Junio C Hamano <gitster@pobox.com>
To: Phillip Wood <phillip.wood123@gmail.com>
Cc: git@vger.kernel.org, Elijah Newren <newren@gmail.com>,
Johannes Sixt <j6t@kdbg.org>
Subject: Re: [PATCH v3 2/2] merge: remember conflict labels
Date: Fri, 09 Oct 2026 17:24:06 -0700 [thread overview]
Message-ID: <xmqq1p9yjtcp.fsf@gitster.g> (raw)
In-Reply-To: <182edb2e8874986206a38a0af005e4df9ef9dd8f.1791537203.git.phillip.wood@dunelm.org.uk> (Phillip Wood's message of "Fri, 9 Oct 2026 10:13:25 +0100")
Phillip Wood <phillip.wood123@gmail.com> writes:
> @@ -347,10 +349,19 @@ static int checkout_merged(int pos, const struct checkout *state,
>
> repo_config_get_bool(the_repository, "merge.renormalize", &renormalize);
> ll_opts.renormalize = renormalize;
> + if (read_merge_labels(the_repository, &base_label, &ours_label,
> + &theirs_label)) {
> + base_label = xstrdup("base");
> + ours_label = xstrdup("ours");
> + theirs_label = xstrdup("theirs");
> + }
> ll_opts.conflict_style = conflict_style;
> - merge_status = ll_merge(&result_buf, path, &ancestor, "base",
> - &ours, "ours", &theirs, "theirs",
> + merge_status = ll_merge(&result_buf, path, &ancestor, base_label,
> + &ours, ours_label, &theirs, theirs_label,
> state->istate, &ll_opts);
> + free(base_label);
> + free(ours_label);
> + free(theirs_label);
> free(ancestor.ptr);
> free(ours.ptr);
> free(theirs.ptr);
This function is called once for each conflicted path, which means
we would read the "merge.renormalize" configuration variable and the
MERGE_LABELS file, both of which will stay constant during a single
conflicted "checkout -m". The issue is shared with the original,
but looking up the same configuration variable repeatedly would be
helped with in-core configset cache. Compared to that, the overhead
added by this patch is to open the same unchanging file, read & parse,
allocate and deallocate.
Perhaps we want to have another preliminary [PATCH 1.5/2] before
this step to allow setting these "per invocation constants" once to
be reused? Then step [PATCH 2/2] can read labels in the prepare
phace just once, use it from the data structure in
checkout_merged(), and free them in release phase when we are done.
builtin/checkout.c | 37 +++++++++++++++++++++++++++++++------
1 file changed, 31 insertions(+), 6 deletions(-)
diff --git i/builtin/checkout.c w/builtin/checkout.c
index c0f0d2c700..a02cbacf1d 100644
--- i/builtin/checkout.c
+++ w/builtin/checkout.c
@@ -310,9 +310,27 @@ static int checkout_stage(int stage, const struct cache_entry *ce, int pos,
return error(_("path '%s' does not have their version"), ce->name);
}
+struct checkout_merged_data {
+ int conflict_style;
+ int renormalize;
+ /* we will add more later */
+};
+
+static void checkout_merged_release(struct checkout_merged_data *data)
+{
+ ; /* nothing to free (yet) */
+}
+
+static void checkout_merged_prepare(struct checkout_merged_data *data)
+{
+ int renormalize = 0;
+ repo_config_get_bool(the_repository, "merge.renormalize", &renormalize);
+ data->renormalize = renormalize;
+}
+
static int checkout_merged(int pos, const struct checkout *state,
int *nr_checkouts, struct mem_pool *ce_mem_pool,
- int conflict_style)
+ struct checkout_merged_data *data)
{
struct cache_entry *ce = the_repository->index->cache[pos];
const char *path = ce->name;
@@ -324,7 +342,6 @@ static int checkout_merged(int pos, const struct checkout *state,
struct object_id threeway[3];
unsigned mode = 0;
struct ll_merge_options ll_opts = LL_MERGE_OPTIONS_INIT;
- int renormalize = 0;
memset(threeway, 0, sizeof(threeway));
while (pos < the_repository->index->cache_nr) {
@@ -345,9 +362,8 @@ static int checkout_merged(int pos, const struct checkout *state,
read_mmblob(&ours, the_repository->objects, &threeway[1]);
read_mmblob(&theirs, the_repository->objects, &threeway[2]);
- repo_config_get_bool(the_repository, "merge.renormalize", &renormalize);
- ll_opts.renormalize = renormalize;
- ll_opts.conflict_style = conflict_style;
+ ll_opts.renormalize = data->renormalize;
+ ll_opts.conflict_style = data->conflict_style;
merge_status = ll_merge(&result_buf, path, &ancestor, "base",
&ours, "ours", &theirs, "theirs",
state->istate, &ll_opts);
@@ -446,6 +462,7 @@ static int checkout_worktree(const struct checkout_opts *opts,
int pos;
int pc_workers, pc_threshold;
struct mem_pool ce_mem_pool;
+ struct checkout_merged_data checkout_merged_data = {0};
state.force = 1;
state.refresh_cache = 1;
@@ -462,6 +479,10 @@ static int checkout_worktree(const struct checkout_opts *opts,
if (pc_workers > 1)
init_parallel_checkout();
+ if (opts->merge) {
+ checkout_merged_prepare(&checkout_merged_data);
+ checkout_merged_data.conflict_style = opts->conflict_style;
+ }
for (pos = 0; pos < the_repository->index->cache_nr; pos++) {
struct cache_entry *ce = the_repository->index->cache[pos];
if (ce->ce_flags & CE_MATCHED) {
@@ -479,10 +500,14 @@ static int checkout_worktree(const struct checkout_opts *opts,
errs |= checkout_merged(pos, &state,
&nr_unmerged,
&ce_mem_pool,
- opts->conflict_style);
+ &checkout_merged_data);
pos = skip_same_name(ce, pos) - 1;
}
}
+
+ if (opts->merge)
+ checkout_merged_release(&checkout_merged_data);
+
if (pc_workers > 1)
errs |= run_parallel_checkout(&state, pc_workers, pc_threshold,
NULL, NULL);
next prev parent reply other threads:[~2026-10-10 0:24 UTC|newest]
Thread overview: 27+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-09-30 9:48 [PATCH 0/2] checkout -m: recreate conflict labels Phillip Wood
2026-09-30 9:48 ` [PATCH 1/2] remove_branch_state: convert boolean argument to flags Phillip Wood
2026-09-30 9:48 ` [PATCH 2/2] merge: remember conflict labels Phillip Wood
2026-09-30 16:42 ` Junio C Hamano
2026-10-01 8:54 ` Phillip Wood
2026-10-01 17:14 ` Junio C Hamano
2026-09-30 20:24 ` [PATCH 0/2] checkout -m: recreate " Johannes Sixt
2026-09-30 20:40 ` Junio C Hamano
2026-09-30 21:20 ` Johannes Sixt
2026-10-01 8:45 ` Phillip Wood
2026-10-05 13:24 ` [PATCH v2 " Phillip Wood
2026-10-05 13:24 ` [PATCH v2 1/2] remove_branch_state: convert boolean argument to flags Phillip Wood
2026-10-05 13:24 ` [PATCH v2 2/2] merge: remember conflict labels Phillip Wood
2026-10-05 16:19 ` Junio C Hamano
2026-10-06 15:21 ` Phillip Wood
2026-10-05 16:31 ` Junio C Hamano
2026-10-06 15:05 ` Phillip Wood
2026-10-06 15:46 ` Junio C Hamano
2026-10-07 13:38 ` Phillip Wood
2026-10-05 14:48 ` [PATCH v2 0/2] checkout -m: recreate " Johannes Sixt
2026-10-05 15:09 ` Phillip Wood
2026-10-05 15:53 ` Junio C Hamano
2026-10-09 9:13 ` [PATCH v3 " Phillip Wood
2026-10-09 9:13 ` [PATCH v3 1/2] remove_branch_state: convert boolean argument to flags Phillip Wood
2026-10-09 9:13 ` [PATCH v3 2/2] merge: remember conflict labels Phillip Wood
2026-10-10 0:24 ` Junio C Hamano [this message]
2026-10-09 20:31 ` [PATCH v3 0/2] checkout -m: recreate " Junio C Hamano
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=xmqq1p9yjtcp.fsf@gitster.g \
--to=gitster@pobox.com \
--cc=git@vger.kernel.org \
--cc=j6t@kdbg.org \
--cc=newren@gmail.com \
--cc=phillip.wood123@gmail.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox