bug#57400: 29.0.50; Support sending patches from VC directly

[Philip Kaludercic]

Robert Pluim <rpluim@gmail.com> writes:

>>>>>> On Wed, 05 Oct 2022 17:34:22 +0000, Philip Kaludercic
> <philipk@posteo.net> said:
>
>     Philip> +@code{vc-prepare-patch} command.  This will prompt you
>     Philip> which revisions
>     Philip> +you wish to share and who the addressee is.  The command
>     Philip> will then use
>     Philip> +your @abbr{MUA, Mail User Agent} for you to review and send out.
>     Philip> +
>
> How about
>
> --begin--
> This will prompt you for the revisions you wish to share, and which
> destination email address(es) to use.  The command will then prepare
> those revisions using your @abbr{MUA, Mail User Agent} for you to
> review and send.
> --end--

I like it.

> The semantics is 'one-or-more addresses', right?

Yes, right.

>     Philip> +@vindex vc-prepare-patches-inline
>     Philip> +Depending on configuration of the user option
>
> "Depending on the value of the user option"

Sounds better.


>     Philip> +@code{vc-prepare-patches-inline}, @code{vc-prepare-patch}
>     Philip> will either
>     Philip> +generate a single or multiple messages.  A @code{nil}
>     Philip> value (the default)
>     Philip> +will prepare and display a message for each revision, one after
>     Philip> +another.  A non-@code{nil} value will have all patches
>     Philip> attached to the
>     Philip> +body of a single message.
>     Philip> +
>
> --begin--
> @code{vc-prepare-patches-inline}, @code{vc-prepare-patch} will
> generate one or more messages.  The default value @code{nil} means
> prepare and display a message for each revision, one after another.  A
> non-@code{nil} value means to generate a single message with all
> patches attached in the body.
> --end--

Sounds a lot better.

>     Philip> +@vindex vc-default-patch-addressee
>     Philip> +If you expect to contribute patches on a regular basis, you can 
> set
>     Philip> +the user option @code{vc-default-patch-addressee} to the address 
> you
>     Philip> +wish to use.  This will be used as the default value when 
> invoking
>     Philip> +@code{vc-prepare-patch}.  Project maintainers may consider 
> setting
>     Philip> +this as a directory local variable (@pxref{Directory Variables}).
>     Philip> +
>
> This can contain multiple addresses, I think, in which case it should
> say so.

Would replacing "address" with "address(es)" suffice?

>     Philip> +** Subr-x
>     Philip> +
>     Philip> +---
>     Philip> +*** New macro 'with-funcall-substitutions'.
>     Philip> +The macro can be used to generically substitute function symbols 
> in
>     Philip> +expressions.
>     Philip> +
>     Philip>  ** Ansi-color
>
> That sounds interesting, but I donʼt see it in the patch :-)

As mentioned in another response, this is from a different patch I hope
to submit soon.  I just had it lying around in etc/NEWS.

>     Philip> +(defcustom vc-prepare-patches-inline nil
>     Philip> +  "Non-nil means that `vc-prepare-patch' creates a single
>     Philip> message.
>
> "Whether `vc-prepare-patch' attaches all revision in a single message."
>
> Iʼm not sure this should have the suffix '-inline', because you can
> have inline attachments and attached attachments, but itʼs not a big
> deal.

If you have a better name, there is no better time to change it than now.

> I also wonder about the default. Creating 100 mail buffers by accident
> is harder to recover from than a single one with 100 attachments, but
> I guess experience will inform us.

The only case where this might happen by accident is when someone
invokes `vc-prepare-patch' in a log-edit buffer where all (or at least a
lot) of revisions have been marked.  In that case, one could add a
"safely check" and make sure that the user actually wants to proceed.

>     Philip> +A single message is created by attaching all patches to the body
>     Philip> +of a single message.  If nil, each patch will be sent out in a
>     Philip> +separate message, which will be prepared sequentially."
>     Philip> +  :type 'boolean
>     Philip> +  :safe #'booleanp
>     Philip> +  :version "29.1")
>     Philip> +
>
> (I didnʼt check, can this do the [PATCH n/m] stuff with the
> subject that 'git format-patch' can do?)

Yes, as the Git backend just copies the subject name that
git-format-patch generates.

>     Philip> +(defcustom vc-default-patch-addressee nil
>     Philip> +  "Default addressee for `vc-prepare-patch'.
>     Philip> +If nil, no default will be used.  This option may be set 
> locally."
>     Philip> +  :type '(choice (const :tag "No default" nil) string)
>     Philip> +  :safe #'stringp
>     Philip> +  :version "29.1")
>     Philip> +
>
> Again, I think this can be multiple addresses. Either as a string
> with commas or as a list of strings perhaps? 

As this is just the default value for `read-multiple-choice' a list with
commae should do.  That being said, how common is it to have multiple
people you consistently want to send a patch to?  Usually you'd have a
central mailing list or something like that, I'd assume.

>     Philip> +;;;###autoload
>     Philip> +(defun vc-prepare-patch (addressee subject revisions)
>     Philip> +  "Compose an Email sending patches for REVISIONS to ADDRESSEE.
>     Philip> +If `vc-prepare-patches-inline' is non-nil, SUBJECT will be used
>     Philip> +as the default subject for the message.  Otherwise a separate
>     Philip> +message will be composed for each revision.
>     Philip> +
>
> ? What does `vc-prepare-patches-inline' have to do with the SUBJECT?

Because the subject for an "inline patch" is extracted from the commit
message.

>     Philip> It includes
>
>     Philip> - some documentation for the Emacs manual and etc/NEWS,
>
>     Philip> - a revised "prepare-patch" interface that uses buffers instead of
>     Philip>   temporary files (I hope this improves the encoding issue),
>
> If itʼs all buffers now then I think you need to update this comment:
>
> +;;
> +;; - prepare-patch (rev)
> +;;
> +;;   Prepare a patch and return a property list with the keys
> +;;   `:subject' indicating the patch message as a string, `:body'
> +;;   containing the contents of the patch as a string (excluding the
> +;;   header) and `:filename' pointing to a file where the patch has
> +;;   been stored.

You are right, thanks for the reminder!

> I have no firm opinion on if there should be a default binding nor
> what it should be 😺
>
> Thanks for this, it will be useful

I'm glad to hear that.  Here is the updated patch:

0001-Add-a-VC-command-to-prepare-patches.patch
Description: Text Data