patchWhy do we need [PATCH]?

Hi git's.

I've just applied a patch to my git tree using git-applymbox. It worked like a charm as it often do. But then when I did a quick check with gitk it stuck me that the subject was prefixed with [PATCH]. I am aware that the [PATCH] in subject tell me: "patch is received from somewhere and applied by me to the .git tree". This is in comparsion to patches received when I merge a git tree for example.

But with common practice to use sign-off I wonder the value of this
patch marker.
When browsing the kernel shortlog I often focus on first word in subject
- cause this tells me what system/drivers is changed. But with the
  [PATCH] marker I have to read some non-sense to see actual subject.

The information is redundant since metadata already tell me who is the author and who committed the change. So could we have it removed or if people continue to find it usefull then at least hide it behind some option. Using the [PATCH] prefix is not the natural thing to do with git.

I looked at the source and found the -k option, but adding [PATCH] should not be default behaviour so this is not the correct solution.

	Sam

On Fri, Dec 16, 2005 at 10:01:45PM +0100, Sam Ravnborg wrote:

 > I've just applied a patch to my git tree using git-applymbox.
 > It worked like a charm as it often do.
 > But then when I did a quick check with gitk it stuck me that
 > the subject was prefixed with [PATCH].
 > I am aware that the [PATCH] in subject tell me:
 > "patch is received from somewhere and applied by me to the .git tree".
 > This is in comparsion to patches received when I merge a git tree for
 > example.
 > 
 > But with common practice to use sign-off I wonder the value of this
 > patch marker.
 > When browsing the kernel shortlog I often focus on first word in subject
 > - cause this tells me what system/drivers is changed. But with the
 >   [PATCH] marker I have to read some non-sense to see actual subject.
 > 
 > The information is redundant since metadata already tell me who is the
 > author and who committed the change.
 > So could we have it removed or if people continue to find it usefull
 > then at least hide it behind some option. Using the [PATCH] prefix is
 > not the natural thing to do with git.
 > 
 > I looked at the source and found the -k option, but adding [PATCH]
 > should not be default behaviour so this is not the correct solution.

I'm still using Linus' original 'applypatch' script, on which I think this stuff was based, and this behaviour annoyed me too. So I changed it to do this..

if [ -f .treename ]; then
  TREENAME=`cat .treename`
else
  TREENAME='PATCH'
fi
(echo "[$TREENAME] $SUBJECT" ; if [ -s $MSGFILE ]; then echo ; cat $MSGFILE; fi ) > $final

and then I have a .treename in say the AGP tree, containing just 'AGPGART', and the changelogs get the prefixes that actually mean something useful. (And those markers are invaluable when searching through history for changes)

		Dave

Sam Ravnborg wrote:

[PATCH] makes sense for the actual email message so that the recipient knows that the message contains a patch. But I agree that it does not add any value to the commit message in git. It actually detracts value, because it takes up limited space in the shortlog message.

--
				Brian Gerst

Sam Ravnborg <sam@ravnborg.org> writes:

The "correct" solution depends on where you come from. That extra [PATCH] is a carryover from BK days, I was told by Linus, to make e-mailed things stand out --- I've never used BK but I am guessing that things were not as obvious as our commit messages, perhaps? We have "Author/Committer" distinction so the [PATCH] marker is redundant.

The rewrite, "git-am" does not bother with adding [PATCH], but the original "git-applymbox", being everyday Linus' tool, was left as it was, not to disrupt the workflow of Linus.

On Fri, Dec 16, 2005 at 01:27:33PM -0800, Junio C Hamano wrote:

You can a typical bk commit here: http://linus.bkbits.net:8080/linux-2.5/cset@1.2243?nav=index.html|ChangeSet@-9M

And here it made much more sense sine the author/comitter info are less structured. Also we do nto have the Signed-off-by: stuff back then.

Thanks, I had forgotten the git-am rewrite. I will use that in the future - if I remember.

	Sam