Subject: Re: Please stop change logs in CVS commit messages
To: NetBSD Packages Technical Discussion List <tech-pkg@netbsd.org>
From: Greg A. Woods <woods@weird.com>
List: tech-pkg
Date: 07/12/2001 15:59:08
[ On Wednesday, July 11, 2001 at 12:10:20 (-0400), Todd Vierling wrote: ]
> Subject: Re: Please stop change logs in CVS commit messages
>
> The pkg-CHANGES commit messages are hardly useful.  If you care about
> bleeding-edge software, you should be reading pkgsrc-changes anyway.  8-)
> 
> (Yes, that's a little tongue-in-cheek.)

Yeah, well, strictly I don't care about bleeding-edge pkgsrc too much,
and I don't subscribe to pkgsrc-changes, and Alistair's wonderful
summary posting here is usualy more than enough information for me.

However I get the doc/pkg-CHANGES commit messages anyway because I read
source-changes (and I suspect I'm not alone here ;-).  Many of its
commit messages have been totally useless either because they contain
way too much junk, or way too little; though this has been improving
somewhat.  I started out just killing all with "CVS commit: doc" as the
subject, but that dropped some of the ones in the same category that I
do care about, and so now I have to skim them all to see which are
which.  I thought about "complaining" about the pkgsrc related stuff
still being on source-changes, but then the quality of doc/pkg-CHANGES
commit messages seemed to improve and so I kept quite.  However they are
still far from being consistent in quality!

So, if doc/pkg-CHANGES is going to stay in doc, and thus it's commit
messages are going to go to what I suspect is a much wider distribution
of people via source-changes, then I humbly repeat my request that they
be very carefully edited to meet these specifications:

	- include the primary CATEGORY/PACKAGE name

	- include the "vendor" version identifier (perhaps combined with
	  the above, so perhaps in terms of makefile variables the first
	  value from CATEGORIES and the value of DISTNAME, separated by
	  a slash, with DISTNAME expanded if it contains other variable
	  references)

	- include the one-line description of the package (i.e. the
	  content of the COMMENT variable from the Makefile)

	- include the PR#(s) if the update/addition was submitted via
	  GNATS, or was prompted by a bug report to GNATS.

	- if the package is an update to an existing package then the
	  message must also include only the new NEWS items from a GNU
	  Coding Standards "NEWS" file (or the equivalent from any other
	  similar "vendor" announcement)

If there's any consensus at all on this I'd even go so far as to suggest
that a special RCS template be put in place for the "doc" module via an
entry such as this in CVSROOT/rcsinfo:

	^doc		$CVSROOT/CVSROOT/rcstemplate.doc

and that template file might contain something like:

CVS: 
CVS: WARNING: You are commiting a change to the NetBSD source repository.
CVS:          This change will be immediately available to all other users
CVS:          of this repository!  Please be sure your changes have been
CVS:          adequately tested and/or proofread.
CVS: 
CVS: NOTICE: By the act of making this commit you are assigning your
CVS: copyright for these changes The NetBSD Foundation.
CVS: 
CVS: If this change is to the pkg-CHANGES file and is in reference to an
CVS: addition or update of a package then please use the following format:
CVS: 
CVS:	{update,addition} of MAIN_CATEGORY/DISTNAME [from PR#NNNNN]
CVS:	
CVS:	Text of the pkgsrc Makefile COMMENT variable.
CVS:	
CVS:	[new items listed in the package's NEWS file]
CVS:	
CVS: [[ etc.  --- perhaps also describe what the pkg-CHANGES entry should
CVS: look like too... ]]


It might even be helpful to have a little script that updates
doc/pkg-CHANGES and commits it all programmatically.  I think that would
be relatively easy given the current pkg-CHANGES format....

-- 
							Greg A. Woods

+1 416 218-0098      VE3TCP      <gwoods@acm.org>     <woods@robohack.ca>
Planix, Inc. <woods@planix.com>;   Secrets of the Weird <woods@weird.com>