Behaviour of log --patch --follow

[Matt Mackall]

On Wed, 2015-05-13 at 23:34 +0100, ekhumoro wrote:
> On 13/05/15 22:47, Matt Mackall wrote:
> > On Wed, 2015-05-13 at 21:36 +0100, ekhumoro wrote:
> >> Hello
> >>
> >> With mercurial version 3.3.3 and earlier, I could use the following to
> >> get a diff for a specific revision that follows history across copies
> >> and renames:
> >>
> >>       hg log --patch --follow -r 233
> >
> > Can you show us an example of what you mean? As written, the above line
> > (with no filenames) says:
> >
> > 1. show revision 233 only
> > 2. show history relative to the working copy
> > 3. show a patch for each revision
> >
> > Up until 3.4, 1 and 2 were basically contradictory and the combination
> > was undefined but we didn't report an error. And 3 is independent: it'll
> > show the same patches for the same files regardless of --follow. By
> > default (without --git), those patches will not report copies and
> > renames (because the standard patch format is deficient in this area)..
> > but will still show the affected files.
> >
> > In 3.4, we changed the behavior of the previously nonsensical --follow
> > --rev (again, with no file) to the useful "follow history from the
> > specified rev (rather than the working copy)" (in other words,
> > equivalent to -r ::233 rather than -r ::.).
> >
> > (All of the above analysis is irrelevant if there's actually a filename
> > in your command you're not telling us about.)
> >
> 
> There's no filename.
> 
> I think the issue here is probably my lack of knowledge. Quite a while 
> ago, I created an alias which does essentially the same as the command 
> above. It solved some problem with renaming (the details of which 
> completely escape me now), and I have been using it regularly ever 
> since. The alias is:
> 
>      fdiff = log --follow --patch
> 
> Which I currently use like this:
> 
>      hg fdiff -r 233
> 
> to see a diff of all the changes made to all the files for a specific 
> revision. Am I wrong to assume that --follow could have ever made any 
> useful difference for this use case?

It's totally unnecessary for that case.. and prior to 3.4, may actually
cause random things to happen.

-- 
Mathematics is the supreme nostalgia of our time.