bug#1537: diff-mode for git patches trouble

bug#1537: diff-mode for git patches trouble

[Dmitriy Blinov]

There is the problem then I'm trying to edit git-generated patches in 
emacs diff-mode.

git patches look like:
---
hello.txt |    2 +-
1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/hello.txt b/hello.txt
index 55bf894..25a09cc 100644
--- a/hello.txt
+++ b/hello.txt
@@ -6,7 +6,7 @@
5
6
7
-8
+8 8
9
10
11
-- 
1.5.6.5


The last but one line is two minuses by default. And diff-mode 
recognizes  first minus as  deleted  string.
Therefore  when we change some strings in patch, emacs creates malformed 
hunk offset.
It will be @@ -6,8 +6,7 @@ insead of correct @@ -6,7 +6,7 @@ after patch 
edition (if we don't delete/create strings).

Is it possible to fix this harm or I should append whitespace before 
"--" string  always?

bug#1537: diff-mode for git patches trouble

[Stefan Monnier]

> The last but one line is two minuses by default.  And diff-mode
> recognizes first minus as deleted string.  Therefore when we change
> some strings in patch, emacs creates malformed hunk offset.  It will
> be @@ -6,8 +6,7 @@ insead of correct @@ -6,7 +6,7 @@ after patch
> edition (if we don't delete/create strings).

> Is it possible to fix this harm or I should append whitespace before "--"
> string always?

Given the way the code is currently implemented, it's difficult to fix
this problem.  One of the problem is that we currently don't do anything
special before the user modifies the buffer.  Not even in
a before-change-function.  So by the time we try to update the hunk
header, the buffer has been modified and we don't know what it looked
like before.
I.e. to fix it we'll need to mark the end-of-hunk before the buffer is
really modified.  E.g. by automatically adding an empty line at
end-of-hunk (or a marker).

Of course, in the general case the problem is insoluble (e.g. if you
edit the line with the "--"), but clearly we can do better in the
"normal" case.


        Stefan

bug#1537: diff-mode for git patches trouble

[Dmitriy Blinov]

 
> Of course, in the general case the problem is insoluble (e.g. if you
> edit the line with the "--"), but clearly we can do better in the
> "normal" case.
>
>
>         Stefan
Probably it was incorrect to post this feature here.  In fact it's git 
problem but not emacs.

bug#1537: diff-mode for git patches trouble

[Stefan Monnier]

severity 1537 wishlist
thanks

>> Of course, in the general case the problem is insoluble (e.g. if you
>> edit the line with the "--"), but clearly we can do better in the
>> "normal" case.
> Probably it was incorrect to post this feature here.  In fact it's git
> problem but not emacs.

No, it definitely has its place here: while it's true that Git's output
is not very friendly to patch-editing as supported by diff-mode, the
output is formally correct and unambiguous.
And your example of changes is also unambiguous, so you can definitely
argue that Emacs should behave correctly in this case.


        Stefan

Processed: Re: bug#1537: diff-mode for git patches trouble

[Emacs bug Tracking System]

Processing commands for control@emacsbugs.donarmstrong.com:

> severity 1537 wishlist
bug#1537: diff-mode for git patches trouble
bug#1536: diff-mode for git patches problem
Severity set to `wishlist' from `normal'

> thanks
Stopping processing here.

Please contact me if you need assistance.

Don Armstrong
(administrator, Emacs bugs database)

bug#1536: bug#1537: diff-mode for git patches trouble

[Stefan Kangas]

Dmitriy Blinov <bda@mns.spb.ru> writes:

> There is the problem then I'm trying to edit git-generated patches in
> emacs diff-mode.
>
> git patches look like:
> ---
> hello.txt |    2 +-
> 1 files changed, 1 insertions(+), 1 deletions(-)
>
> diff --git a/hello.txt b/hello.txt
> index 55bf894..25a09cc 100644
> --- a/hello.txt
> +++ b/hello.txt
> @@ -6,7 +6,7 @@
> 5
> 6
> 7
> -8
> +8 8
> 9
> 10
> 11

(That was 11 years ago.)

I can't reproduce this on current master.  The numbers seem to update
correctly, as ar as I'm understanding from:

    https://stackoverflow.com/questions/6508651/what-do-the-numbers-on-a-git-diff-header-mean

It would be useful if someone could double check this finding in case
there is something I'm missing.  It could be quickly verified using:

    git format-patch -1 ; ./src/emacs -Q *patch

And then edit any line in a patch to see if the numbers update correctly.

Best regards,
Stefan Kangas

bug#1537: bug#1536: bug#1537: diff-mode for git patches trouble

[Lars Ingebrigtsen]

Stefan Kangas <stefan@marxist.se> writes:

> I can't reproduce this on current master.  The numbers seem to update
> correctly, as ar as I'm understanding from:
>
>     https://stackoverflow.com/questions/6508651/what-do-the-numbers-on-a-git-diff-header-mean
>
> It would be useful if someone could double check this finding in case
> there is something I'm missing.  It could be quickly verified using:
>
>     git format-patch -1 ; ./src/emacs -Q *patch
>
> And then edit any line in a patch to see if the numbers update correctly.

Yup; seems to update correctly for me, too, so I'm closing this bug report.

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no