[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: thoughts about svnpatch

From: Stefan Sperling <stsp_at_elego.de>
Date: Mon, 22 Jun 2009 20:37:49 +0100

On Mon, Jun 22, 2009 at 09:09:15PM +0200, Stefan Kng wrote:
> Hi,
> I'm currently playing around with the new svnpatch feature. I haven't
> got it working quite yet though, but I'd like to propose some
> changes/enhancements:


> * applying an svnpatch file only works if the patch file is applied to a
> working copy. Would it be possible to make this work for unversioned
> source trees too?

Should be possible to do. Please open an issue.

> * I'm missing a dry-run function to preview what applying the patch
> would do - there's a 'dry-run' variable in the patch code, so I guess
> this is already planned?

I hadn't planned this yet.
What exactly would you like this feature to do?
Should it just send notifications?

Note that the 'svn patch' code consists of two parts:

1. The part which concerns itself with svnpatch blocks. These blocks
   basically contain compressed svn:// protocol information encoded
   in ASCII text. This was written by Charles Acknin.
2. The part which concerns itself with applying unidiff files.
   This part was mostly written by myself, and replaced the
   call to an external patch program which was originally
   part of part 1.

There's a comment in libsvn_client/patch.c which separates the
two parts:

  Part 1
/* --- Text-diff application routines --- */
  Part 2

Since I didn't know about the dry-run variable, I guess it belongs
to part 1. I don't have much of an idea about part 1, expect that
it looks to me to add far too much new code for what it's doing
and I think we should try to find a way to trim down the size of
part 1 before releasing it. We can release both parts independently
of one another.

In case you're interested (and would like to help :) this is what
still remains to be done for part 2 before we can release it:

- detect eol-style of the patch file and the target file,
  and use the appropriate eol indicator when reading from the patch
  file and writing to the target -- we want patches generated on windows
  to work on *nix and vice-versa.
- line offset searching so that we can avoid merge conflicts
  if the target file contains the original hunk text at a
  different line than the original file.
- a strip-count option (-p) for stripping leading directory
  components from target paths specified in the patch file
- a --encoding option for specifying the patch file's encoding in case
  it is not encoded in the charset of the currently active locale
- a few other TODOs in libsvn_client/patch.c, mostly handling error
I haven't opened issues for any of these, maybe I should...

> * I would also like to have some kind of filter for svn_client_patch so
> only items which match the filter will actually get patched. That way I
> can present the user with a list of affected items (with the dry-run
> feature) and then let the user choose which of the items (s)he actually
> wants to patch.

Good idea! Please open an issue for this, too.

> * does the svn_client_patch() only work with svnpatch files or also with
> unified diff files? From browsing through the code I think it should?

It can and always has worked on any unidiff files. svnpatch blocks
are ignored as noise by standard patch tools.

Received on 2009-06-22 21:38:10 CEST

This is an archived mail posted to the Subversion Dev mailing list.

This site is subject to the Apache Privacy Policy and the Apache Public Forum Archive Policy.