Subject: Re: Patch naming
To: None <>
From: Ben Collver <>
List: tech-pkg
Date: 12/30/2003 20:22:30
On Tue, Dec 30, 2003 at 08:37:00PM -0500, Chuck Yerkes wrote:
> Hmmm, trying to get this to work with db3 rather than db2.  
> Okay, did that...  now build it.  Doh, it must be patching configure.
> Lets derive which file it is...

You point here isn't too clear.

> "It's just how we've done it" is often a bad answer.
> There is no "win" by using "patch-aa"  vs "patch-configure".
> "it's prettier" and "it makes them line up in neat columns" and
> "I can quickly count them"  are about all I can come up with.
> And they're pretty flaccid reasons.

The Debian package system just has a single file containing all the
differences from the official program distribution.  For the purpose
of revision control, NetBSD uses a separate patch file for each file that
is changed.  I suspect this is the good reason to have multiple patch
files, and patch-?? is a convenient way to have multiple file names.

I would prefer games/falcons-eye/patches/patch-bi over an OpenBSD style
games/falcons-eye/patches/patch-win_jtp_gamedata_config_jtp_keys_txt any

> If there's a choice between doing it good and doing it better and the
> effort [not totally minor] is "follow the new form as you upgrade
> patches", then there's not a lot of reason to embrace
> obtuseness|mediocrity.  That's Microsoft's job.

You're the one saying the current system is mediocre.  The Microsoft
comparison is completely useless.

> what does (random grab?) /usr/pkgsrc/devel/cmdline/patches/patch-ak
> change?  Oh, 2 levels of abstraction to find out

Would it be even less obtuse if we added a list of the modified functions
or classes in the patch filename?

A hacker does for love what others would not do for money.