Re: [Quilt-dev] Does anyone know why ...

[Peter Williams]

Andreas Gruenbacher wrote:
> On Thursday, 02 March 2006 07:25, Peter Williams wrote:
>
> > ... "quilt files <patch>" produces empty output when <patch> is not
> > applied?  Was this a deliberate design decision?  If so what was the
> > logic behind it?
>
>
> The current version of quilt will tell you that the specified patch is not 
> applied; it doesn't fail silently.
>
> Quilt won't try to list the files that an unapplied patch touches because this 
> can't be done very well: GNU patch uses complicated heuristics to decide 
> which filename it will patch

This shouldn't be necessary for patches in quilt.  It should be VERY 
well known which files will be effected by a patch.

> (see Multiple Patches in a File in the diff info 
> pages). You can use utilities like lsdiff for that purpose,

That involves me knowing/exploiting the internal workings of quilt which 
is not a "good thing"(TM) IMHO.

> but lsdiff will 
> also guess wrong for some patches.

If I remember correctly Andrew's patch management utilities which were 
the inspiration for quilt (I believe) kept the list of files in a patch 
as a separate piece of data in the database.  Apart from the fact it 
would enable "quilt files <patch>" to give a useful response, doing 
something similar in quilt would also cure one of its other annoyances 
namely files going missing from a patch if the patch is popped when 
there is currently no modifications to the file.

Although this problem could be described as a feature (i.e. automatic 
trimming of the patch file list) I think that it's a bug.  It can lead 
to problems such as the user editing a file because he thinks it's in 
the patch because he put it there and didn't take it out.

Peter
PS I also think that "patch diff" should display diffs for unapplied 
patches IFF there is a patch file available.  It's sometimes very useful 
to be able to check what changes a patch will make before it is applied. 
 And yes, I know that I could look at the file directly BUT that 
involves me knowing/exploiting the internal workings of quilt which is 
not a "good thing"(TM).
--
Peter Williams                                   address@hidden

"Learning, n. The kind of ignorance distinguishing the studious."
 -- Ambrose Bierce