Re: [PATCH v2 2/2] merge: remember conflict labels
On 06/10/2026 16:46, Junio C Hamano wrote:
Show 16 quoted lines
> Phillip Wood <phillip.wood123@gmail.com> writes:
> >> @@ -4969,6 +4973,13 @@ void merge_switch_to_result(struct merge_options *opt,
>> return;
>> }
>> trace2_region_leave("merge", "write_auto_merge", opt->repo);
>> +
>> + trace2_region_enter("merge", "write_merge_labels", opt->repo);
>> + opt->priv = result->priv;
>> + write_merge_labels(opt->repo, opt->priv->labels[0], opt->priv->labels[1],
>> + opt->priv->labels[2]);
>> + opt->priv = NULL;
>> + trace2_region_leave("merge", "write_merge_labels", opt->repo);
> > A "if (result->clean >= 0 && update_worktree_and_index)" condition
> guards this code path, so we unconditionall call
> write_merge_labels(), whether the result is clean or with conflict.
> Am I reading the code correctly?Yes, I think so. Ouch! how did I miss that?
Will fix, thanks
Phillip
Show 30 quoted lines
>> @@ -5234,6 +5245,14 @@ static void move_opt_priv_to_result_priv(struct merge_options *opt,
>> * to move it.
>> */
>> assert(opt->priv && !result->priv);
>> + if (!result->clean) {
>> + opt->priv->labels[0] =
>> + mem_pool_strdup(&opt->priv->pool, opt->ancestor);
>> + opt->priv->labels[1] =
>> + mem_pool_strdup(&opt->priv->pool, opt->branch1);
>> + opt->priv->labels[2] =
>> + mem_pool_strdup(&opt->priv->pool, opt->branch2);
>> + }
> > But we only populate the labels[] when conflicted. What would we
> write when we do not have conflicts?
> >> +int write_merge_labels(struct repository *r, const char *base,
>> + const char *ours, const char *theirs)
>> +{
>> + FILE *f = fopen_or_warn(git_path_merge_labels(r), "w");
>> +
>> + if (!f)
>> + return -1;
>> +
>> + fprintf(f, "%s\n%s\n%s\n", base, ours, theirs);
>> + if (fclose(f))
>> + return error_errno("could not write '%s'",
>> + git_path_merge_labels(r));
>> +
>> + return 0;
>> +}
> > Would the answer be "(null)\n(null)\n(null)\n"?