[PATCH] status: fix verbose status coloring inconsistency

[PATCH] status: fix verbose status coloring inconsistency

[Lance Ward via GitGitGadget]

From: Lance Ward <ljward10@gmail.com>

Currently setting color.status=always results in a colored diff when
going stdout, but an uncolored diff when going to other files or piping
to other commands such as less or more.  This patch fixes this and now
color.status=always implies color.diff=always regardless of the output
location.

Signed-off-by: Lance Ward <ljward10@gmail.com>
---
    status: fix verbose status coloring inconsistency
    
    Currently setting color.status=always results in a colored diff when
    going stdout, but an uncolored diff when going to other files or piping
    to other commands such as less or more. This patch fixes this and now
    color.status=always implies color.diff=always regardless of the output
    location.
    
    Signed-off-by: Lance Ward ljward10@gmail.com

Published-As: https://github.com/gitgitgadget/git/releases/tag/pr-git-954%2Fljward10%2Flw-fix-status-coloring-v1
Fetch-It-Via: git fetch https://github.com/gitgitgadget/git pr-git-954/ljward10/lw-fix-status-coloring-v1
Pull-Request: https://github.com/git/git/pull/954

 t/t7527-status-color-pipe.sh | 55 ++++++++++++++++++++++++++++++++++++
 wt-status.c                  |  2 ++
 2 files changed, 57 insertions(+)
 create mode 100755 t/t7527-status-color-pipe.sh

diff --git a/t/t7527-status-color-pipe.sh b/t/t7527-status-color-pipe.sh
new file mode 100755
index 00000000000..88df03ae066
--- /dev/null
+++ b/t/t7527-status-color-pipe.sh
@@ -0,0 +1,55 @@
+#!/bin/sh
+
+test_description='git status color option'
+
+. ./test-lib.sh
+
+test_expect_success setup '
+	echo 1 >original &&
+	git add .
+'
+
+# Normal git status does not pipe colors
+test_expect_success 'git status' '
+	git status >raw &&
+	test_decode_color <raw >out &&
+	grep "original$" out
+'
+
+# Test color.status=never (expect same as above)
+test_expect_success 'git -c color.status=never status' '
+	git -c color.status=never status >raw &&
+	test_decode_color <raw >out &&
+	grep "original$" out
+'
+
+# Test color.status=always
+test_expect_success 'git -c color.status=always status' '
+	git -c color.status=always status >raw &&
+	test_decode_color <raw >out &&
+	grep "original<RESET>$" out
+'
+
+# Test verbose (default)
+test_expect_success 'git status -v' '
+	git status -v >raw &&
+	test_decode_color <raw >out &&
+	grep "+1" out
+'
+
+# Test verbose color.status=never
+test_expect_success 'git -c color.status=never status -v' '
+	git -c color.status=never status -v >raw &&
+	test_decode_color <raw >out &&
+	grep "+1" out
+'
+
+# Test verbose color.status=always
+test_expect_success 'git -c color.status=always status -v' '
+	git -c color.status=always status -v >raw &&
+	test_decode_color <raw >out &&
+	grep "<CYAN>@@ -0,0 +1 @@<RESET>" out &&
+	grep "GREEN>+<RESET><GREEN>1<RESET>" out
+'
+
+test_done
diff --git a/wt-status.c b/wt-status.c
index 0c8287a023e..1e9c899a7b2 100644
--- a/wt-status.c
+++ b/wt-status.c
@@ -1064,6 +1064,8 @@ static void wt_longstatus_print_verbose(struct wt_status *s)
 	if (s->fp != stdout) {
 		rev.diffopt.use_color = 0;
 		wt_status_add_cut_line(s->fp);
+	} else {
+		rev.diffopt.use_color = s->use_color;
 	}
 	if (s->verbose > 1 && s->committable) {
 		/* print_updated() printed a header, so do we */

base-commit: e6362826a0409539642a5738db61827e5978e2e4
-- 
gitgitgadget

Re: [PATCH] status: fix verbose status coloring inconsistency

[Junio C Hamano]

"Lance Ward via GitGitGadget" <gitgitgadget@gmail.com> writes:

> From: Lance Ward <ljward10@gmail.com>
>
> Currently setting color.status=always results in a colored diff when

Our log message begins with the description of the current status,
so "Currently" is not something you need to say.

What command with what options?  "git status" does not even show
"diff" at least by default. 

    ... goes, experiments, and guesses what the poster means ...

Perhaps you meant something like this:

    status: honor color.status=always when sending diff output to non tty

    "git status --verbose" shows the patch in color by default (in
    addition to the list of added, modified, and/or deleted paths)
    when the output goes to a tty.  With color.status configuration
    set to 'always' and sending the output to a non-tty, the list of
    paths are colored as expected, but the patch output loses colors.

And then after the description of the current status, you give
orders to a patch monkey to fix the code "to be like so".

    This is because the code did not pass the settings read from the
    configuration and the command line to the underlying machinery
    that generates the patch output.  Fix it to do so.

> going stdout, but an uncolored diff when going to other files or piping
> to other commands such as less or more.  This patch fixes this and now
> color.status=always implies color.diff=always regardless of the output
> location.
>
> Signed-off-by: Lance Ward <ljward10@gmail.com>
> ---

Eric may deserve Helped-by: mention.

>  t/t7527-status-color-pipe.sh | 55 ++++++++++++++++++++++++++++++++++++

Don't we already have test that checks "status" output including its
coloring already? I'd rather see us adding to existing test script,
than allocating a new number for a small subset of features being
tested for a command.  After all, test numbers are limited resource.

> +test_expect_success setup '
> +	echo 1 >original &&
> +	git add .
> +'
> +
> +# Normal git status does not pipe colors

What does "pipe colors" mean?  "color its output on a non-tty", you mean?

> +test_expect_success 'git status' '
> +	git status >raw &&
> +	test_decode_color <raw >out &&
> +	grep "original$" out
> +'

Not "new file: *original$" or something less false-positive prone?

> +# Test color.status=never (expect same as above)
> +test_expect_success 'git -c color.status=never status' '
> +	git -c color.status=never status >raw &&
> +	test_decode_color <raw >out &&
> +	grep "original$" out
> +'
> +

Would it make sense to have tests for color.status=true, I wonder.
It requires tty to actually "see" the colors output but sending
the output to tty is the normal use case, so we should care...

> +# Test color.status=always
> +test_expect_success 'git -c color.status=always status' '
> +	git -c color.status=always status >raw &&
> +	test_decode_color <raw >out &&
> +	grep "original<RESET>$" out
> +'

OK.  I understand that this passes without the patch below.

> +# Test verbose (default)
> +test_expect_success 'git status -v' '
> +	git status -v >raw &&
> +	test_decode_color <raw >out &&
> +	grep "+1" out
> +'

I think you meant to catch the new contents "1" stored in the file
whose name is "original", but this will hit the hunk header, no?

    @@ -0,0 +1 @@
    +1

IOW, the grep patterns throughout this patch may be a bit too loose
and prone to false hits.

> +# Test verbose color.status=never
> +test_expect_success 'git -c color.status=never status -v' '
> +	git -c color.status=never status -v >raw &&
> +	test_decode_color <raw >out &&
> +	grep "+1" out
> +'
> +
> +# Test verbose color.status=always
> +test_expect_success 'git -c color.status=always status -v' '
> +	git -c color.status=always status -v >raw &&
> +	test_decode_color <raw >out &&
> +	grep "<CYAN>@@ -0,0 +1 @@<RESET>" out &&
> +	grep "GREEN>+<RESET><GREEN>1<RESET>" out
> +'

Is the missing open bra "<" before GREEN intended?

Are we forcing the standard palette?

> +test_done
> diff --git a/wt-status.c b/wt-status.c
> index 0c8287a023e..1e9c899a7b2 100644
> --- a/wt-status.c
> +++ b/wt-status.c
> @@ -1064,6 +1064,8 @@ static void wt_longstatus_print_verbose(struct wt_status *s)
>  	if (s->fp != stdout) {
>  		rev.diffopt.use_color = 0;
>  		wt_status_add_cut_line(s->fp);
> +	} else {
> +		rev.diffopt.use_color = s->use_color;
>  	}
>  	if (s->verbose > 1 && s->committable) {
>  		/* print_updated() printed a header, so do we */
>
> base-commit: e6362826a0409539642a5738db61827e5978e2e4

Thanks.

Re: [PATCH] status: fix verbose status coloring inconsistency

[Lance Ward]

On Wed, Feb 3, 2021 at 4:51 PM Junio C Hamano <gitster@pobox.com> wrote:
>
> "Lance Ward via GitGitGadget" <gitgitgadget@gmail.com> writes:
>
> > From: Lance Ward <ljward10@gmail.com>
> >
> > Currently setting color.status=always results in a colored diff when
>
> Our log message begins with the description of the current status,
> so "Currently" is not something you need to say.
>
> What command with what options?  "git status" does not even show
> "diff" at least by default.
>
>     ... goes, experiments, and guesses what the poster means ...

I'm disappointed by your tone...

I'll go ahead and close my pull requests, if someone else wants
to pick them up that's fine with me.

>
> Perhaps you meant something like this:
>
>     status: honor color.status=always when sending diff output to non tty
>
>     "git status --verbose" shows the patch in color by default (in
>     addition to the list of added, modified, and/or deleted paths)
>     when the output goes to a tty.  With color.status configuration
>     set to 'always' and sending the output to a non-tty, the list of
>     paths are colored as expected, but the patch output loses colors.
>
> And then after the description of the current status, you give
> orders to a patch monkey to fix the code "to be like so".
>
>     This is because the code did not pass the settings read from the
>     configuration and the command line to the underlying machinery
>     that generates the patch output.  Fix it to do so.
>
> > going stdout, but an uncolored diff when going to other files or piping
> > to other commands such as less or more.  This patch fixes this and now
> > color.status=always implies color.diff=always regardless of the output
> > location.
> >
> > Signed-off-by: Lance Ward <ljward10@gmail.com>
> > ---
>
> Eric may deserve Helped-by: mention.
>
> >  t/t7527-status-color-pipe.sh | 55 ++++++++++++++++++++++++++++++++++++
>
> Don't we already have test that checks "status" output including its
> coloring already? I'd rather see us adding to existing test script,
> than allocating a new number for a small subset of features being
> tested for a command.  After all, test numbers are limited resource.
>
> > +test_expect_success setup '
> > +     echo 1 >original &&
> > +     git add .
> > +'
> > +
> > +# Normal git status does not pipe colors
>
> What does "pipe colors" mean?  "color its output on a non-tty", you mean?
>
> > +test_expect_success 'git status' '
> > +     git status >raw &&
> > +     test_decode_color <raw >out &&
> > +     grep "original$" out
> > +'
>
> Not "new file: *original$" or something less false-positive prone?
>
> > +# Test color.status=never (expect same as above)
> > +test_expect_success 'git -c color.status=never status' '
> > +     git -c color.status=never status >raw &&
> > +     test_decode_color <raw >out &&
> > +     grep "original$" out
> > +'
> > +
>
> Would it make sense to have tests for color.status=true, I wonder.
> It requires tty to actually "see" the colors output but sending
> the output to tty is the normal use case, so we should care...
>
> > +# Test color.status=always
> > +test_expect_success 'git -c color.status=always status' '
> > +     git -c color.status=always status >raw &&
> > +     test_decode_color <raw >out &&
> > +     grep "original<RESET>$" out
> > +'
>
> OK.  I understand that this passes without the patch below.
>
> > +# Test verbose (default)
> > +test_expect_success 'git status -v' '
> > +     git status -v >raw &&
> > +     test_decode_color <raw >out &&
> > +     grep "+1" out
> > +'
>
> I think you meant to catch the new contents "1" stored in the file
> whose name is "original", but this will hit the hunk header, no?
>
>     @@ -0,0 +1 @@
>     +1
>
> IOW, the grep patterns throughout this patch may be a bit too loose
> and prone to false hits.
>
> > +# Test verbose color.status=never
> > +test_expect_success 'git -c color.status=never status -v' '
> > +     git -c color.status=never status -v >raw &&
> > +     test_decode_color <raw >out &&
> > +     grep "+1" out
> > +'
> > +
> > +# Test verbose color.status=always
> > +test_expect_success 'git -c color.status=always status -v' '
> > +     git -c color.status=always status -v >raw &&
> > +     test_decode_color <raw >out &&
> > +     grep "<CYAN>@@ -0,0 +1 @@<RESET>" out &&
> > +     grep "GREEN>+<RESET><GREEN>1<RESET>" out
> > +'
>
> Is the missing open bra "<" before GREEN intended?
>
> Are we forcing the standard palette?
>
> > +test_done
> > diff --git a/wt-status.c b/wt-status.c
> > index 0c8287a023e..1e9c899a7b2 100644
> > --- a/wt-status.c
> > +++ b/wt-status.c
> > @@ -1064,6 +1064,8 @@ static void wt_longstatus_print_verbose(struct wt_status *s)
> >       if (s->fp != stdout) {
> >               rev.diffopt.use_color = 0;
> >               wt_status_add_cut_line(s->fp);
> > +     } else {
> > +             rev.diffopt.use_color = s->use_color;
> >       }
> >       if (s->verbose > 1 && s->committable) {
> >               /* print_updated() printed a header, so do we */
> >
> > base-commit: e6362826a0409539642a5738db61827e5978e2e4
>
> Thanks.

Re: [PATCH] status: fix verbose status coloring inconsistency

[Eric Sunshine]

On Wed, Feb 3, 2021 at 7:44 PM Lance Ward <ljward10@gmail.com> wrote:
> On Wed, Feb 3, 2021 at 4:51 PM Junio C Hamano <gitster@pobox.com> wrote:
> > Our log message begins with the description of the current status,
> > so "Currently" is not something you need to say.
>
> I'm disappointed by your tone...
>
> I'll go ahead and close my pull requests, if someone else wants
> to pick them up that's fine with me.

It is, unfortunately, easy to misinterpret a reviewer's neutral tone
as being negative or as an attempt to shame the author. But be assured
that the goal of reviewers on this project is to help the patch author
get the submission into proper shape for acceptance, and when Junio
takes the time to write such a comprehensive review, he does so
because he sees promise in both the patches and in the author of the
patches. A review as extensive as this one is intended to get the
newcomer up to speed quickly with local project conventions (such as
how commit messages are written) and to help land the patches with as
few revisions as possible since both submitter and reviewer time is
valuable.

Re: [PATCH] status: fix verbose status coloring inconsistency

[Eric Sunshine]

On Wed, Feb 3, 2021 at 5:51 PM Junio C Hamano <gitster@pobox.com> wrote:
> "Lance Ward via GitGitGadget" <gitgitgadget@gmail.com> writes:
> >  t/t7527-status-color-pipe.sh | 55 ++++++++++++++++++++++++++++++++++++
>
> Don't we already have test that checks "status" output including its
> coloring already? I'd rather see us adding to existing test script,
> than allocating a new number for a small subset of features being
> tested for a command.  After all, test numbers are limited resource.

t7508 seems like a good place for these tests.

> > +test_expect_success 'git status' '
> > +     git status >raw &&
> > +     test_decode_color <raw >out &&
> > +     grep "original$" out
> > +'
>
> Not "new file: *original$" or something less false-positive prone?

The "new file:" message is localized, so this `grep` will need to
become `test_i18ngrep` again (as it was in the original submission) if
this approach is adopted, which is fine.

> > +test_expect_success 'git -c color.status=never status' '
> > +     git -c color.status=never status >raw &&
> > +     test_decode_color <raw >out &&
> > +     grep "original$" out
> > +'
>
> Would it make sense to have tests for color.status=true, I wonder.
> It requires tty to actually "see" the colors output but sending
> the output to tty is the normal use case, so we should care...

I wondered if `color.status=true` might already be tested by t7508 but
apparently it only checks `auto`.

> > +test_expect_success 'git -c color.status=always status -v' '
> > +     git -c color.status=always status -v >raw &&
> > +     test_decode_color <raw >out &&
> > +     grep "<CYAN>@@ -0,0 +1 @@<RESET>" out &&
> > +     grep "GREEN>+<RESET><GREEN>1<RESET>" out
> > +'
>
> Are we forcing the standard palette?

As this is a stand-alone test script with a well-controlled initial
configuration, I expect it would be using the default palette. t7508
does assign a custom palette for `color.status` but not, apparently,
for `color.diff`, so it presumably should be okay there, as well.