Re: patch for spelling mistake

On Tue, Jul 10, 2001 at 08:58:30PM -0400, Bruce Momjian wrote:
> > nconway(at)klamath(dot)dyndns(dot)org (Neil Conway) writes:
> > > I think I made the patch properly; from the root of the 'pgsql'
> > > CVS module, apply with: patch -p0 < patchfile.
> >
> > diff -c output is preferred. A plain diff such as you have here has too
> > little context to be safe to apply (if someone else has edited the file
> > since the version you diffed from, a plain diff patch is much too likely
> > to be applied incorrectly).
>
> Tom is right, diff -c is best, though in the patch you applied, there is
> no chance for error so I can accept it. You have unique previous text
> for each change.

Sorry about that -- and thanks for the tip (Tom & Bruce).

I've remade the patch using the make_diff tools -- that seems to
produce the preferred the format. The updated patch is below.

I'm going to add this info to the Developer's FAQ, but it seems like
there are 2 copies -- one plaintext and one HTML: it seems as though
they are updated simultaneously (according to the CVS logs anyway).
Which version should I make changes to? Both? Are they generated from
a single source document somewhere?

Cheers,

Neil

*** ./src/backend/commands/cluster.c.orig Tue Jul 10 21:34:14 2001
--- ./src/backend/commands/cluster.c Tue Jul 10 21:34:30 2001
***************
*** 45,57 ****
* cluster
*
* STILL TO DO:
! * Create a list of all the other indicies on this relation. Because
* the cluster will wreck all the tids, I'll need to destroy bogus
! * indicies. The user will have to re-create them. Not nice, but
* I'm not a nice guy. The alternative is to try some kind of post
* destroy re-build. This may be possible. I'll check out what the
* index create functiond want in the way of paramaters. On the other
! * hand, re-creating n indicies may blow out the space.
*/
void
cluster(char *oldrelname, char *oldindexname)
--- 45,57 ----
* cluster
*
* STILL TO DO:
! * Create a list of all the other indexes on this relation. Because
* the cluster will wreck all the tids, I'll need to destroy bogus
! * indexes. The user will have to re-create them. Not nice, but
* I'm not a nice guy. The alternative is to try some kind of post
* destroy re-build. This may be possible. I'll check out what the
* index create functiond want in the way of paramaters. On the other
! * hand, re-creating n indexes may blow out the space.
*/
void
cluster(char *oldrelname, char *oldindexname)