threads / discuss / 15564
Subject: combined diff, but not condensed, howto?
Johannes SixtAfter a merge conflict, 'git diff' shows a combined diff, which presents only the parts that conflicted or where there are near-by changes from different parents (potential conflicts). Is there a command line switch so that *all* changes are shown, even non-conflicting ones?
'git diff', 'git diff --cc', and 'git diff -c' all show the same output.
-- Hannes
Junio C HamanoRe: combined diff, but not condensed, howto?
Johannes Sixt <j.sixt@viscovery.net> writes:
> After a merge conflict, 'git diff' shows a combined diff, which presents > only the parts that conflicted or where there are near-by changes from > different parents (potential conflicts). Is there a command line switch so > that *all* changes are shown, even non-conflicting ones?
Like "git diff HEAD"?
Johannes SixtRe: combined diff, but not condensed, howto?
Junio C Hamano schrieb:
> Johannes Sixt <j.sixt@viscovery.net> writes: > >> After a merge conflict, 'git diff' shows a combined diff, which presents >> only the parts that conflicted or where there are near-by changes from >> different parents (potential conflicts). Is there a command line switch so >> that *all* changes are shown, even non-conflicting ones? > > Like "git diff HEAD"?
No. Something that produces
@@@ -23,5 -23,5 +23,7 @@@ A B ++<<<<<<< HEAD:foo +C ++======= + D ++>>>>>>> 97d7fee2cb068f215a593c6e5623b265db45d3bc:foo E F @@@ -50,4 -50,5 +53,5 @@@ G H +I J K @@@ -60,5 -61,4 +63,4 @@@ L M - N O P
Junio C HamanoRe: combined diff, but not condensed, howto?
Johannes Sixt <j.sixt@viscovery.net> writes:
> Junio C Hamano schrieb: >> Johannes Sixt <j.sixt@viscovery.net> writes: >> >>> After a merge conflict, 'git diff' shows a combined diff, which presents >>> only the parts that conflicted or where there are near-by changes from >>> different parents (potential conflicts). Is there a command line switch so >>> that *all* changes are shown, even non-conflicting ones? >> >> Like "git diff HEAD"? > > No. Something that produces
...
Ah, I see what you mean.
It all happens inside combine-diff.c::make_hunks(). If you pass dense==0, you should be able to get all the uninteresting hunks, I think.
Perhaps
$ git diff --base -c
???
Johannes SixtRe: combined diff, but not condensed, howto?
Junio C Hamano schrieb:
> Johannes Sixt <j.sixt@viscovery.net> writes: > >> Junio C Hamano schrieb: >>> Johannes Sixt <j.sixt@viscovery.net> writes: >>> >>>> After a merge conflict, 'git diff' shows a combined diff, which presents >>>> only the parts that conflicted or where there are near-by changes from >>>> different parents (potential conflicts). Is there a command line switch so >>>> that *all* changes are shown, even non-conflicting ones? >>> Like "git diff HEAD"? >> No. Something that produces > ... > > Ah, I see what you mean. > > It all happens inside combine-diff.c::make_hunks(). If you pass dense==0, > you should be able to get all the uninteresting hunks, I think. > > Perhaps > > $ git diff --base -c
Yes, that does it!
Thanks, -- Hannes
Junio C HamanoRe: combined diff, but not condensed, howto?
Johannes Sixt <j.sixt@viscovery.net> writes:
> Junio C Hamano schrieb: > ... >> It all happens inside combine-diff.c::make_hunks(). If you pass dense==0, >> you should be able to get all the uninteresting hunks, I think. >> >> Perhaps >> >> $ git diff --base -c > > Yes, that does it!
It may do so, but I do not necessarily agree with the logic.
It is very much Ok and useful for "git diff" to default to "--cc -p", and it may be fine for "git diff-files -p" to default to "--cc".
But I think ideally an explicit "git diff -c" and "git diff-files -c -p" should not turn the "dense combined" mode on.
The commonality of these two functions is striking and calls for proper refactoring, but aside from that, I think a patch like this should be applied. What do you think?
---
builtin-diff-files.c | 7 ++++++- builtin-diff.c | 7 ++++++- 2 files changed, 12 insertions(+), 2 deletions(-)
diff --git i/builtin-diff-files.c w/builtin-diff-files.c index 9bf10bb..2b578c7 100644 --- i/builtin-diff-files.c +++ w/builtin-diff-files.c @@ -50,7 +50,12 @@ int cmd_diff_files(int argc, const char **argv, const char *prefix) 3 < rev.max_count) usage(diff_files_usage); - if (rev.max_count == -1 && + /* + * "diff-files --base -p" should not combine merges because it + * was not asked to. "diff-files -c -p" should not densify + * (the user should ask with "diff-files --cc" explicitly). + */ + if (rev.max_count == -1 && !rev.combine_merges && (rev.diffopt.output_format & DIFF_FORMAT_PATCH)) rev.combine_merges = rev.dense_combined_merges = 1; diff --git i/builtin-diff.c w/builtin-diff.c index 037c303..9fb30c6 100644 --- i/builtin-diff.c +++ w/builtin-diff.c @@ -223,7 +223,12 @@ static int builtin_diff_files(struct rev_info *revs, int argc, const char **argv argv++; argc--; } - if (revs->max_count == -1 && + /* + * "diff --base" should not combine merges because it was not + * asked to. "diff -c" should not densify (the user should ask + * with "diff --cc" explicitly. + */ + if (revs->max_count == -1 && !revs->combine_merges && (revs->diffopt.output_format & DIFF_FORMAT_PATCH)) revs->combine_merges = revs->dense_combined_merges = 1;
Johannes SixtRe: combined diff, but not condensed, howto?
Junio C Hamano schrieb:
> It is very much Ok and useful for "git diff" to default to "--cc -p", and > it may be fine for "git diff-files -p" to default to "--cc". > > But I think ideally an explicit "git diff -c" and "git diff-files -c -p" > should not turn the "dense combined" mode on. > > The commonality of these two functions is striking and calls for proper > refactoring, but aside from that, I think a patch like this should be > applied. What do you think?
This makes a lot of sense. As I mention in my initial post, I was wondering why 'diff -c' was the same as 'diff --cc'.
> diff --git i/builtin-diff-files.c w/builtin-diff-files.c > index 9bf10bb..2b578c7 100644 > --- i/builtin-diff-files.c > +++ w/builtin-diff-files.c > @@ -50,7 +50,12 @@ int cmd_diff_files(int argc, const char **argv, const char *prefix) > 3 < rev.max_count) > usage(diff_files_usage); > > - if (rev.max_count == -1 && > + /* > + * "diff-files --base -p" should not combine merges because it > + * was not asked to. "diff-files -c -p" should not densify > + * (the user should ask with "diff-files --cc" explicitly). > + */ > + if (rev.max_count == -1 && !rev.combine_merges && > (rev.diffopt.output_format & DIFF_FORMAT_PATCH)) > rev.combine_merges = rev.dense_combined_merges = 1;
Hm... In my tests, a bare 'git diff-files --cc' produces a combined diff, but 'git diff-files -c' does not.
> diff --git i/builtin-diff.c w/builtin-diff.c > index 037c303..9fb30c6 100644 > --- i/builtin-diff.c > +++ w/builtin-diff.c > @@ -223,7 +223,12 @@ static int builtin_diff_files(struct rev_info *revs, int argc, const char **argv > argv++; argc--; > } > > - if (revs->max_count == -1 && > + /* > + * "diff --base" should not combine merges because it was not > + * asked to. "diff -c" should not densify (the user should ask > + * with "diff --cc" explicitly.
I don't see why you add "the user should ask with 'diff --cc'" when this becomes the default anyway.
> + */ > + if (revs->max_count == -1 && !revs->combine_merges && > (revs->diffopt.output_format & DIFF_FORMAT_PATCH)) > revs->combine_merges = revs->dense_combined_merges = 1;
-- Hannes
Junio C Hamano[PATCH] diff/diff-files: do not use --cc too aggressively
Textual diff output for unmerged paths was too eager to give condensed combined diff. This has two problems:
(1) "diff --base/--ours/--theirs" (and "diff-files -p" with the same) is
a request to compare with the named stage. We showed "-c -p" output
instead; (2) "diff -c" (and "diff-files -c -p") is a request to view combined diff
without condensing (otherwise the user would have explicitly asked
for --cc, not -c). We showed "--cc" output anyway.0fe7c1d (built-in diff: assorted updates, 2006-04-29) claimed to be careful about doing this, but its breakage was hidden because back then "git diff" was still a shell script that did not use the codepath it introduced fully.
Signed-off-by: Junio C Hamano <gitster@pobox.com> --- builtin-diff-files.c | 7 ++++++- builtin-diff.c | 8 +++++++- 2 files changed, 13 insertions(+), 2 deletions(-)
diff --git a/builtin-diff-files.c b/builtin-diff-files.c index 9bf10bb..2b578c7 100644 --- a/builtin-diff-files.c +++ b/builtin-diff-files.c @@ -50,7 +50,12 @@ int cmd_diff_files(int argc, const char **argv, const char *prefix) 3 < rev.max_count) usage(diff_files_usage); - if (rev.max_count == -1 && + /* + * "diff-files --base -p" should not combine merges because it + * was not asked to. "diff-files -c -p" should not densify + * (the user should ask with "diff-files --cc" explicitly). + */ + if (rev.max_count == -1 && !rev.combine_merges && (rev.diffopt.output_format & DIFF_FORMAT_PATCH)) rev.combine_merges = rev.dense_combined_merges = 1; diff --git a/builtin-diff.c b/builtin-diff.c index 037c303..d5fe775 100644 --- a/builtin-diff.c +++ b/builtin-diff.c @@ -223,7 +223,13 @@ static int builtin_diff_files(struct rev_info *revs, int argc, const char **argv argv++; argc--; } - if (revs->max_count == -1 && + /* + * "diff --base" should not combine merges because it was not + * asked to. "diff -c" should not densify (if the user wants + * dense one, --cc can be explicitly asked for, or just rely + * on the default). + */ + if (revs->max_count == -1 && !revs->combine_merges && (revs->diffopt.output_format & DIFF_FORMAT_PATCH)) revs->combine_merges = revs->dense_combined_merges = 1;
-- 1.6.0.2.412.g0bc80
Johannes SixtRe: [PATCH] diff/diff-files: do not use --cc too aggressively
Junio C Hamano schrieb:
> Textual diff output for unmerged paths was too eager to give condensed > combined diff. This has two problems: > > (1) "diff --base/--ours/--theirs" (and "diff-files -p" with the same) is > a request to compare with the named stage. We showed "-c -p" output > instead;
I cannot reproduce what you describe here. "diff --base/--ours/--theirs" works as expected with and without this patch.
> (2) "diff -c" (and "diff-files -c -p") is a request to view combined diff > without condensing (otherwise the user would have explicitly asked > for --cc, not -c). We showed "--cc" output anyway.
The patch fixes this:
Tested-by: Johannes Sixt <johannes.sixt@telecom.at>
-- Hannes