Jiang Xin <worldhello.net@xxxxxxxxx> writes:
> When pusing with SSH or other smart protocol, references are validated
> by function `check_to_send_update()` beforce they are sent in commands
beforce -> before
> to `send_pack()` of "receve-pack". For atomic push, if a reference is
> rejected after the validation, only references pushed by user should be
> marked as failure, instead of report failure on all remote references.
>
> Commit v2.22.0-1-g3bca1e7f9f (transport-helper: enforce atomic in
> push_refs_with_push, 2019-07-11) wanted to fix report issue of HTTP
> protocol, but marked all remote references failure for atomic push.
>
> Revert part of that commit and add additional status for function
> `atomic_push_failure()`. The additional status for it except the
> "REF_STATUS_EXPECTING_REPORT" status are:
>
> - REF_STATUS_NONE : Not marked as "REF_STATUS_EXPECTING_REPORT" yet.
> - REF_STATUS_OK : Assume OK for dryrun or status_report is disabled.
>
> This commit break test case in t5541, and will fix in other commit.
Presumably the fix in 3/3 is too big to be included in here? Let's
see how it goes, as such a strategy sometimes helps understand the
change and sometimes does not.
> diff --git a/send-pack.c b/send-pack.c
> index 0407841ae8..ff016c468c 100644
> --- a/send-pack.c
> +++ b/send-pack.c
> @@ -334,6 +334,8 @@ static int atomic_push_failure(struct send_pack_args *args,
> continue;
>
> switch (ref->status) {
> + case REF_STATUS_NONE:
> + case REF_STATUS_OK:
> case REF_STATUS_EXPECTING_REPORT:
> ref->status = REF_STATUS_ATOMIC_PUSH_FAILED;
> continue;
This is fed the whole *remote_refs list, iterates over them but we
only come here if the ref is actually pushed (i.e. peer_ref is set,
or we are doing a mirror push), so it does sound like the right
place to do this.
