how to `register' vc-hg-region-history in vc-region-history?

how to `register' vc-hg-region-history in vc-region-history?

[Uwe Brauer]

Hi

I just tried out a new function
(right now I don't care much about keybinding and font locking

So the function 
(defun vc-hg-region-history (file buffer lfrom lto) 

Seems a natural generalization of vc-git-region-history  However I 
don't see how to tell vc-region-history to tell that the new 
function now exists. I tried to debug it


(defun vc-region-history (from to) 
    (vc-call region-history file buf lfrom lto) 

And there vc aborts for HG. What do I miss?

Uwe Brauer 

Re: how to `register' vc-hg-region-history in vc-region-history?

[Dmitry Gutov]

On 1/18/18 17:27, Uwe Brauer wrote:
> 
> Hi
> 
> I just tried out a new function
> (right now I don't care much about keybinding and font locking
> 
> So the function (defun vc-hg-region-history (file buffer lfrom lto)
> Seems a natural generalization of vc-git-region-history  However I don't 
> see how to tell vc-region-history to tell that the new function now 
> exists. I tried to debug it
> 
> 
> (defun vc-region-history (from to)    (vc-call region-history file buf 
> lfrom lto)
> And there vc aborts for HG. What do I miss?

You also need vc-hg-region-history-mode. vc-region-history calls it.

Re: how to `register' vc-hg-region-history in vc-region-history?

[Stefan Monnier]

Uwe Brauer <oub@mat.ucm.es> writes:
> Seems a natural generalization of vc-git-region-history  However I don't see
> how to tell vc-region-history to tell that the new function now
> exists.

VC automatically discovers that.  But it also remembers what it failed
to find, os if you had tried vc-region-history in the same session
before defining that function, you need to

    (put 'HG 'vc-functions nil)

to flush VC's cache of known backend functions.
Not sure why this line is missing from vc-hg.el, contrarily to, say, vc-cvs.el.


        Stefan

Re: how to `register' vc-hg-region-history in vc-region-history?

[Stefan Monnier]

>     (put 'HG 'vc-functions nil)
           ^^^
           'Hg

I plan on installing the patch below into master as soon as
git.sv.gnu.org responds again,


        Stefan


diff --git a/lisp/vc/vc-git.el b/lisp/vc/vc-git.el
index 626cf6165a..2d6ca1386a 100644
--- a/lisp/vc/vc-git.el
+++ b/lisp/vc/vc-git.el
@@ -183,6 +183,10 @@ vc-git-log-output-coding-system
 ;; History of Git commands.
 (defvar vc-git-history nil)
 
+;; Clear up the cache to force vc-call to check again and discover
+;; new functions when we reload this file.
+(put 'Git 'vc-functions nil)
+
 ;;; BACKEND PROPERTIES
 
 (defun vc-git-revision-granularity () 'repository)
diff --git a/lisp/vc/vc-hg.el b/lisp/vc/vc-hg.el
index 08b1be8f6d..ad817fd9b9 100644
--- a/lisp/vc/vc-hg.el
+++ b/lisp/vc/vc-hg.el
@@ -175,6 +175,10 @@ vc-hg-root-log-format
   :version "24.5")
 
 \f
+;; Clear up the cache to force vc-call to check again and discover
+;; new functions when we reload this file.
+(put 'Hg 'vc-functions nil)
+
 ;;; Properties of the backend
 
 (defvar vc-hg-history nil)

Re: how to `register' vc-hg-region-history in vc-region-history?

[Uwe Brauer]

>>> "Stefan" == Stefan Monnier <monnier@iro.umontreal.ca> writes:

   > Uwe Brauer <oub@mat.ucm.es> writes:
   >> Seems a natural generalization of vc-git-region-history  However I don't see
   >> how to tell vc-region-history to tell that the new function now
   >> exists.

   > VC automatically discovers that.  But it also remembers what it failed
   > to find, os if you had tried vc-region-history in the same session
   > before defining that function, you need to

   >     (put 'HG 'vc-functions nil)

That is what I was looking for.

I can test my function now.

But alas, I don't understand enough of git.

The git function has the following call


(vc-call-backend 'hg 'diff file "HEAD" nil (current-buffer))


Mercurial has no HEAD as part of a command. Would that be tip for
mercurial? What do the vc-hg.el authors say?

Re: how to `register' vc-hg-region-history in vc-region-history?

[Stefan Monnier]

> (vc-call-backend 'hg 'diff file "HEAD" nil (current-buffer))

The purpose of this call is to get the diff between the current file's
contents and the contents of the version in which "git log -L" will
interpret the line numbers.

> Mercurial has no HEAD as part of a command.  Would that be tip for
> mercurial?

Probably, but it depends on what version is used by `hg` when it
interprets the line numbers of `hg log -fp -L foo.c,10:20`.


        Stefan