tabs, git diff and git apply

tabs, git diff and git apply

[martin rudalics]

Some three lines appearing in my window.c as

       if (WINDOWP (w->contents))
	cont = foreach_window_1 (XWINDOW (w->contents), fn, user_data);
       else

appear in *shell* after git diff as

+      if (WINDOWP (w->contents))
+ 	cont = foreach_window_1 (XWINDOW (w->contents), fn, user_data);
+      else

Note the extra space after the "+" in the second line and the off-by-one
indentation.  A subsequent git apply then complains about

/home/martin/temp/master..changes-2020-11-28.diff:3847: space before tab in indent.
  	cont = foreach_window_1 (XWINDOW (w->contents), fn, user_data);

Now my sources contain many lines where text is preceded by exactly one
tab on its line but this is the only one that produces such a failure.
Does anyone have an explanation for that behavior?

Thanks in advance, martin

Re: tabs, git diff and git apply

[Andreas Schwab]

What does git status tell?  How did you call git diff?

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 7578 EB47 D4E5 4D69 2510  2552 DF73 E780 A9DA AEC1
"And now for something completely different."

Re: tabs, git diff and git apply

[martin rudalics]

 > What does git status tell?

You probably mean git status after git apply.  It just lists the changed
files and doesn't care about the whitespace problem.

But I didn't cite one line (to avoid confusions due to my German locale)
emitted by git apply:

warning: 1 Zeile fügt Whitespace-Fehler hinzu.

 > How did you call git diff?

git diff master..changes | tee

where changes is the branch where my changes sit.

martin

Re: tabs, git diff and git apply

[Andreas Schwab]

On Nov 28 2020, martin rudalics wrote:

>> What does git status tell?
>
> You probably mean git status after git apply.

No, before.

> It just lists the changed files and doesn't care about the whitespace
> problem.

This is not useful.  Be precise.  Don't leave anything out.

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 7578 EB47 D4E5 4D69 2510  2552 DF73 E780 A9DA AEC1
"And now for something completely different."

Re: tabs, git diff and git apply

[martin rudalics]

 >>> What does git status tell?
 >>
 >> You probably mean git status after git apply.
 >
 > No, before.
 >
 >> It just lists the changed files and doesn't care about the whitespace
 >> problem.
 >
 > This is not useful.  Be precise.  Don't leave anything out.

The entire sequence is:

martin@restno:~/emacs-git/trunk$ git status
Auf Branch master
Ihr Branch ist auf demselben Stand wie 'origin/master'.

Unversionierte Dateien:
   (benutzen Sie "git add <Datei>...", um die Änderungen zum Commit vorzumerken)

	obj-gtk-2/
	obj-gtk/
	obj-lucid/
	obj-motif/
	obj-none/
	obj-ns/
	obj-opt/
	obj-wo/

nichts zum Commit vorgemerkt, aber es gibt unversionierte Dateien
(benutzen Sie "git add" zum Versionieren)
martin@restno:~/emacs-git/trunk$ git apply /home/martin/temp/master..changes-2020-11-28.diff
/home/martin/temp/master..changes-2020-11-28.diff:3847: space before tab in indent.
  	cont = foreach_window_1 (XWINDOW (w->contents), fn, user_data);
warning: 1 Zeile fügt Whitespace-Fehler hinzu.
martin@restno:~/emacs-git/trunk$ git status
Auf Branch master
Ihr Branch ist auf demselben Stand wie 'origin/master'.

Änderungen, die nicht zum Commit vorgemerkt sind:
   (benutzen Sie "git add <Datei>...", um die Änderungen zum Commit vorzumerken)
   (benutzen Sie "git checkout -- <Datei>...", um die Änderungen im Arbeitsverzeichnis zu verwerfen)

	geändert:       lisp/cus-start.el
	geändert:       lisp/face-remap.el
	geändert:       lisp/mouse.el
	geändert:       lisp/simple.el
	geändert:       lisp/window.el
	geändert:       src/alloc.c
	geändert:       src/buffer.c
	geändert:       src/buffer.h
	geändert:       src/data.c
	geändert:       src/dispextern.h
	geändert:       src/dispnew.c
	geändert:       src/eval.c
	geändert:       src/font.c
	geändert:       src/frame.c
	geändert:       src/frame.h
	geändert:       src/indent.c
	geändert:       src/keyboard.c
	geändert:       src/lisp.h
	geändert:       src/nsterm.m
	geändert:       src/w32fns.c
	geändert:       src/w32term.c
	geändert:       src/widget.c
	geändert:       src/window.c
	geändert:       src/window.h
	geändert:       src/xdisp.c
	geändert:       src/xfaces.c
	geändert:       src/xmenu.c
	geändert:       src/xterm.c

Unversionierte Dateien:
   (benutzen Sie "git add <Datei>...", um die Änderungen zum Commit vorzumerken)

	obj-gtk-2/
	obj-gtk/
	obj-lucid/
	obj-motif/
	obj-none/
	obj-ns/
	obj-opt/
	obj-wo/

keine Änderungen zum Commit vorgemerkt (benutzen Sie "git add" und/oder "git commit -a")
martin@restno:~/emacs-git/trunk$

Re: tabs, git diff and git apply

[Andreas Schwab]

On Nov 28 2020, martin rudalics wrote:

> martin@restno:~/emacs-git/trunk$ git apply /home/martin/temp/master..changes-2020-11-28.diff
> /home/martin/temp/master..changes-2020-11-28.diff:3847: space before tab in indent.
>  	cont = foreach_window_1 (XWINDOW (w->contents), fn, user_data);
> warning: 1 Zeile fügt Whitespace-Fehler hinzu.

So you are applying a patch with whitespace errors.  What is the issue?

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 7578 EB47 D4E5 4D69 2510  2552 DF73 E780 A9DA AEC1
"And now for something completely different."

Re: tabs, git diff and git apply

[martin rudalics]

 > So you are applying a patch with whitespace errors.  What is the issue?

That that "whitespace error" was introduced by git diff.  The character
at the beginning of that line in window.c

	cont = foreach_window_1 (XWINDOW (w->contents), fn, user_data);

is described by Emacs as


              position: 8315 of 341150 (2%), column: 0
             character: TAB (displayed as TAB) (codepoint 9, #o11, #x9)
               charset: ascii (ASCII (ISO646 IRV))
code point in charset: 0x09
                script: latin
                syntax:   	which means: whitespace
              to input: type "C-x 8 RET 9" or "C-x 8 RET CHARACTER TABULATION"
           buffer code: #x09
             file code: #x09 (encoded by coding system undecided-unix)
               display: no font available

Character code properties: customize what to show
   old-name: CHARACTER TABULATION
   general-category: Cc (Other, Control)

There are text properties here:
   fontified            t


Now where does the space produced by git diff come from so that in

+ 	cont = foreach_window_1 (XWINDOW (w->contents), fn, user_data);

the character following the "+" is described by Emacs as


              position: 163661 of 384033 (43%), column: 1
             character: SPC (displayed as SPC) (codepoint 32, #o40, #x20)
               charset: ascii (ASCII (ISO646 IRV))
code point in charset: 0x20
                script: latin
                syntax:   	which means: whitespace
              category: .:Base, a:ASCII, l:Latin
              to input: type "C-x 8 RET 20" or "C-x 8 RET SPACE"
           buffer code: #x20
             file code: #x20 (encoded by coding system utf-8-unix)
               display: by this font (glyph code)
     ftcrhb:-PfEd-DejaVu Sans Mono-normal-normal-normal-*-13-*-*-*-m-0-iso10646-1 (#x03)
        hardcoded face: nobreak-space

Character code properties: customize what to show
   name: SPACE
   general-category: Zs (Separator, Space)
   decomposition: (32) (' ')

There are 2 overlays here:
  From 163088 to 164494
   diff--font-lock-syntax t
   diff-mode            syntax
   evaporate            t
   modification-hooks   (diff--overlay-auto-delete)
  From 163088 to 164494
   diff--font-lock-refined t
   diff-mode            fine
   evaporate            t
   modification-hooks   (diff--overlay-auto-delete)


There are text properties here:
   face                 diff-added
   fontified            t


martin

Re: tabs, git diff and git apply

[Andreas Schwab]

On Nov 28 2020, martin rudalics wrote:

>> So you are applying a patch with whitespace errors.  What is the issue?
>
> That that "whitespace error" was introduced by git diff.

You didn't do any git diff in your entire sequence.

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 7578 EB47 D4E5 4D69 2510  2552 DF73 E780 A9DA AEC1
"And now for something completely different."

Re: tabs, git diff and git apply

[martin rudalics]

> You didn't do any git diff in your entire sequence.

Right.  It was a change I didn't commit.

Thanks, martin

Re: tabs, git diff and git apply

[Stefan Monnier]

> Some three lines appearing in my window.c as
>
>       if (WINDOWP (w->contents))
> 	cont = foreach_window_1 (XWINDOW (w->contents), fn, user_data);
>       else
>
> appear in *shell* after git diff as
>
> +      if (WINDOWP (w->contents))
> + 	cont = foreach_window_1 (XWINDOW (w->contents), fn, user_data);
> +      else
>
> Note the extra space after the "+" in the second line and the off-by-one
> indentation.

Any chance you're using non-unified diffs?  These match the
"[-+ ] <LINE>" pattern, contrary to unified diffs which use "[-+ ]<LINE>".

IOW, maybe there's nothing wrong.

> A subsequent git apply then complains about
>
> /home/martin/temp/master..changes-2020-11-28.diff:3847: space before tab in indent.
>  	cont = foreach_window_1 (XWINDOW (w->contents), fn, user_data);

Hard to tell without more info.
Have you tried it in an xterm rather than withing a *shell* buffer?


        Stefan

Re: tabs, git diff and git apply

[martin rudalics]

 > Any chance you're using non-unified diffs?  These match the
 > "[-+ ] <LINE>" pattern, contrary to unified diffs which use "[-+ ]<LINE>".

No.  I forgot to commit the fix so diff still found and correctly added
that spurious space at the beginning of the line.  It's OK now.

Thanks, martin

Re: tabs, git diff and git apply

[Stefan Monnier]

>> Any chance you're using non-unified diffs?  These match the
>> "[-+ ] <LINE>" pattern, contrary to unified diffs which use "[-+ ]<LINE>".
> No.  I forgot to commit the fix so diff still found and correctly added
> that spurious space at the beginning of the line.  It's OK now.

I hate those tools which blindly do what we say instead of what we mean!


        Stefan

Re: tabs, git diff and git apply

[Alfred M. Szmidt]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 354 bytes --]

   martin@restno:~/emacs-git/trunk$ git apply /home/martin/temp/master..changes-2020-11-28.diff
   /home/martin/temp/master..changes-2020-11-28.diff:3847: space before tab in indent.
	   cont = foreach_window_1 (XWINDOW (w->contents), fn, user_data);
   warning: 1 Zeile fügt Whitespace-Fehler hinzu.

How was master..changes-2020-11-28.diff created?

Re: tabs, git diff and git apply

[martin rudalics]

 > I hate those tools which blindly do what we say instead of what we mean!

git diff surely has the option needed to do what I meant here.  Just
that for me its nomenclature is worse than Emacs', so I couldn't tell it
what I meant ...

martin

Re: tabs, git diff and git apply

[martin rudalics]

 > How was master..changes-2020-11-28.diff created?

With

git diff master..changes | tee

where "changes" is the name of the branch that contains my changes.  It
would be nice to add an option to warn me about uncommitted changes ...

martin

Re: tabs, git diff and git apply

[Alfred M. Szmidt]

    > How was master..changes-2020-11-28.diff created?

   With

   git diff master..changes | tee

   where "changes" is the name of the branch that contains my changes.  It
   would be nice to add an option to warn me about uncommitted changes ...

If you do that again, do you get the same type of diff? Does `git
diff` have a configuration that is doing context diffs?

This seems to be an issue with `git diff` rather than with emacs?

Re: tabs, git diff and git apply

[martin rudalics]

 > If you do that again, do you get the same type of diff?

I get the same result.

 > Does `git
 > diff` have a configuration that is doing context diffs?

Sure.  You just have to be able to understand the doc.

 > This seems to be an issue with `git diff` rather than with emacs?

It's an issue with me or better with me and git.  I'm only using Emacs to
run git commands from it.

martin

Re: tabs, git diff and git apply

[Alfred M. Szmidt]

    > If you do that again, do you get the same type of diff?

   I get the same result.

    > Does `git
    > diff` have a configuration that is doing context diffs?

   Sure.  You just have to be able to understand the doc.

That is easy to verify, and you should be able to do

   HOME=/tmp git diff

and see if the behaviour still is the same (if it doesn't work, just
running as a different user should).  If it isn't, it is a
configuraiton setting that you have done -- reading the git-config man
page would be the next action to see where such a setting is set.

Re: tabs, git diff and git apply

[martin rudalics]

 >      > If you do that again, do you get the same type of diff?
 >
 >     I get the same result.
 >
 >      > Does `git
 >      > diff` have a configuration that is doing context diffs?
 >
 >     Sure.  You just have to be able to understand the doc.
 >
 > That is easy to verify, and you should be able to do
 >
 >     HOME=/tmp git diff
 >
 > and see if the behaviour still is the same (if it doesn't work, just
 > running as a different user should).  If it isn't, it is a
 > configuraiton setting that you have done -- reading the git-config man
 > page would be the next action to see where such a setting is set.

But that wasn't my problem.  My problem was that I changed a file but
did not commit the change.  So git diff picked up the pre-change file
contents and used them instead of the actual file contents.  When I
looked at the diff output I was stumped by the fact that it claimed the
presence of a SPC character I couldn't find in that file.  The rule is
to always run git status before git diff but I probably had forgotten it
this time.

martin

Re: tabs, git diff and git apply

[Yuri Khan]

On Sun, 29 Nov 2020 at 16:13, Alfred M. Szmidt <ams@gnu.org> wrote:

>     > Does `git
>     > diff` have a configuration that is doing context diffs?
>
>    Sure.  You just have to be able to understand the doc.
>
> That is easy to verify, and you should be able to do
>
>    HOME=/tmp git diff
>
> and see if the behaviour still is the same (if it doesn't work, just
> running as a different user should).  If it isn't, it is a
> configuraiton setting that you have done -- reading the git-config man
> page would be the next action to see where such a setting is set.

Don’t bother. Git diff always does a unified diff.