Re: b4: introducing b4 shazam (like b4 am -o- | git am)

[Konstantin Ryabitsev <konstantin@linuxfoundation.org>]

On Wed, Sep 22, 2021 at 02:38:03PM -0300, Jason Gunthorpe wrote:
> > - you can then "git checkout -b foo FETCH_HEAD" or "git merge FETCH_HEAD"
> >   or "git rebase" as necessary
> 
> Does it prep the merge message too? ie pull in the text from the cover
> letter, make the ==== markers/etc? (ie look what davem does:
> d1bf73387b5adbc12c6a59c1fbaa69e05ee265ed)

We could, theoretically, do that, but you'd have to remember to pass the -F
--edit flags to "git merge". My concern is mostly that cover letters are very
much free-form and there isn't a sane approach we can take to automatically
generate good merge commit messages from them, short of quoting them fully,
which will be overkill 9 times out of 10.

One thing I will probably do is edit .git/FETCH_HEAD to change the unhelpful
/tmp/tmpfoobar to say something like:

    [sha]\t\tpatch series from https://lore.kernel.org/r/msgid

This way the default "git merge" message will contain a clickable link for the
person doing the merge where they can see the cover letter and maybe
copy-paste from it.

> > Alternatively, if you run "git shazam -A", we'll just run "git am" on the
> > current HEAD, which is the exact equivalent of "b4 am -o- | git am".
> 
> It might be interesting to detect which to do based on what the
> submitter has done.. No useful base-commit = no merge?

This is where I have to defer to feedback. I expect it would depend on the
size of the series -- for smaller ones people will probably want to rebase and
not merge?

> I'm generally of the feeling that git is a VCS and it is completely
> fine for a tool to do whatever to the tree because I can quickly undo
> it. So I think b4 is possibly overly cautious here :)

Well, doing it via FETCH_HEAD was also an easy way to use base-commit info
(whether real or deduced) without asking for things like "what would you like
the branch to be named" etc.

> > If the default operation with getting things ready in FETCH_HEAD is not
> > useful, please let me know how you would prefer to see things work instead.
> 
> I would be nice to have some process guidance when maintainers should
> be using merges and when am'd patches..

I'll be happy to consider any feedback here.

-K