bug#18954: 24.4; vc-log Summary: header is included in the log message

[Ivan Shmakov <ivan@siamics.net>]

>>>>> Ed Avis <eda@waniasset.com> writes:

 > I use vc-mode to edit and commit files in an svn repository.  On
 > running vc-next-action on a modified file, a log message buffer
 > appears containing the text Summary: and a blank line.  Typing in
 > this buffer adds more text after Summary:.  On committing with C-c
 > C-c, the literal text Summary: is included as part of the svn log
 > message, as you can see afterwards by running 'svn log'.

 > Please see <http://emacs.stackexchange.com/questions/2717/> where
 > forum user Stefan indicated that this was a bug (presumably, the
 > Summary: header should be stripped out when sending the text to svn)
 > and asked to report it.

	This pseudo-header (along with an optional Author:) is added by
	the log-edit-insert-message-template command, itself called from
	log-edit-hook.  Thus, as a work-around (which I happen to use
	myself), it’s possible to simply remove the function from the
	hook, like:

(eval-after-load 'log-edit
  '(remove-hook 'log-edit-hook 'log-edit-insert-message-template))

	A quick scan through vc/*.el suggest that the Summary: header is
	only recognized by vc-arch.el (which, however, doesn’t seem to
	take care of Author:), while only vc-git.el seems to recognize
	Author: (and doesn’t seem to care about Summary: in turn.)

	If so, this bug actually affects most of the VCS supported by
	the Emacs VC facility, and not just Subversion.

	I believe this behavior is highly inconsistent, and would rather
	suggest either of the following options.

	• Amend the log-edit-insert-message-template function to only
	  generate the template if /either/ explicitly requested by the
	  user (as in: via a yes-maybe-no variable setting, or by
	  running it interactively) /or/ when the headers in question
	  are known to be recognized by the target VCS.

	• Remove the function from the hook’s default value.

	Also to note that log-edit-insert-cvs-rcstemplate apparently
	relies on the existence of CVS/Root to decide if its respective
	template should be added.  My guess is that it should instead
	check the VCS chosen by the user (as in: vc-switch-backend) for
	the fileset in question.

-- 
FSF associate member #7257  http://boycottsystemd.org/  … 3013 B6A0 230E 334A