Is GGG mislabeling topics?

Is GGG mislabeling topics?

[Junio C Hamano]

These comments from GGG bot

    https://github.com/git/git/pull/1435#issuecomment-1386301994
    https://github.com/git/git/pull/1435#issuecomment-1386302018

add 'next' and 'seen' labels, citing merges e3ead5f and c52b021
respectively, but these merges are of a topic that has little to do
with this pull request (#1435).  Is this expected?

Re: Is GGG mislabeling topics?

[Johannes Schindelin]

Hi Junio,

On Thu, 19 Jan 2023, Junio C Hamano wrote:

> These comments from GGG bot
>
>     https://github.com/git/git/pull/1435#issuecomment-1386301994
>     https://github.com/git/git/pull/1435#issuecomment-1386302018
>
> add 'next' and 'seen' labels, citing merges e3ead5f and c52b021
> respectively, but these merges are of a topic that has little to do
> with this pull request (#1435).  Is this expected?

Since I could not make `amlog` work reliably for GitGitGadget (and you
will recall that it was not for lack of trying, not by _any_ stretch of
imagination, I really, really put in a few weeks and it just turned out
that there was no way to come up with a 100% correct record of mappings
between patches on the Git mailing list and the commits in git/git), the
best I could do was to use `range-diff` to identify which patches made it
upstream, and what the upstream OIDs are.

What you see in action is that this is imperfect. Because in the absence
of actual git/git commits that correspond to the GitGitGadget Pull
Requests' commits, `range-diff` will even identify replacements or
alternative patches as the git/git commits corresponding to the PR.

I don't see how this could be made robust without forcing you to adopt
different processes than you are using right now.

Ciao,
Johannes

Re: Is GGG mislabeling topics?

[Junio C Hamano]

Johannes Schindelin <Johannes.Schindelin@gmx.de> writes:

> On Thu, 19 Jan 2023, Junio C Hamano wrote:
>
>> These comments from GGG bot
>>
>>     https://github.com/git/git/pull/1435#issuecomment-1386301994
>>     https://github.com/git/git/pull/1435#issuecomment-1386302018
>>
>> add 'next' and 'seen' labels, citing merges e3ead5f and c52b021
>> respectively, but these merges are of a topic that has little to do
>> with this pull request (#1435).  Is this expected?
>
> Since I could not make `amlog` work reliably for GitGitGadget ...

"git rev-list e3ead5f^..e3ead5f" (as the comment claims that GGG saw
e3ead5f merged this topic) should be able to identify the individual
commits merged, in this case three topmost commits from Peff leading
to 6c065f72 (http: support CURLOPT_PROTOCOLS_STR, 2023-01-16).

> What you see in action is that this is imperfect. Because in the absence
> of actual git/git commits that correspond to the GitGitGadget Pull
> Requests' commits, `range-diff` will even identify replacements or
> alternative patches as the git/git commits corresponding to the PR.

Hmph, so the three topmost commits were mistaken to match the
patch(es) on this branch.  That's sort-of understandable but not
very satisfactory.

And amlog seems to know the message ID that resulted in that topmost
commit that was merged to 'next'.

    $ git notes --ref notes/amlog show e3ead5f^2
    Message-Id: <Y8YQUD8bMHc4Lmph@coredump.intra.peff.net>

Perhaps GGG can be told to recognise that it is not one of the
messages it sent out before, or something?