git diff -U0 header off-by-one error when deleting no lines

git diff -U0 header off-by-one error when deleting no lines

[Lukas Tenbrink]

Thank you for filling out a Git bug report!
Please answer the following questions to help us understand your issue.

What did you do before the bug happened? (Steps to reproduce your issue)

Full reproduction:

echo "A\nB\nC" > test1
echo "A\nB\nC\nD\nE\nF" > test2
git diff -U0 test1 test2


What did you expect to happen? (Expected behavior)

I expected the header of the change to show the correct "from line index".
For reference, a plain git diff:

> git diff test1 test2
diff --git a/test1 b/test2
index b1e6722..cead32e 100644
--- a/test1
+++ b/test2
@@ -1,3 +1,6 @@
 A
 B
 C
+D
+E
+F

When passing -U0, I would expect the from line index to be 3 larger, because the 3 lines of context are omitted:

diff --git a/test1 b/test2
index b1e6722..cead32e 100644
--- a/test1
+++ b/test2
@@ -4,0 +4,3 @@ C
+D
+E
+F


What happened instead? (Actual behavior)

The actual output places the new text at line 3:

> git diff -U0 test1 test2
diff --git a/test1 b/test2
index b1e6722..cead32e 100644
--- a/test1
+++ b/test2
@@ -3,0 +4,3 @@ C
+D
+E
+F


What's different between what you expected and what actually happened?

The header information "-3" is incorrect, because if a script wishes to insert the added code using the line indicator, it will place the newly added code at line 3 (before the C).
Instead, I would expect "-4", because we wish to insert the DEF starting at line 4.


Anything else you want to add:

Notably, the bug does not occur when also deleting lines.
Now, we are correctly placed at the beginning of line 3:

echo "A\nB\nD\nE\nF" > test3

> git diff -U0 test1 test3
diff --git a/test1 b/test3
index b1e6722..98422e9 100644
--- a/test1
+++ b/test3
@@ -3 +3,3 @@ B
-C
+D
+E
+F


[System Info]
git version:
git version 2.39.2 (Apple Git-143)
cpu: x86_64
no commit associated with this build
sizeof-long: 8
sizeof-size_t: 8
shell-path: /bin/sh
feature: fsmonitor--daemon
uname: Darwin 22.4.0 Darwin Kernel Version 22.4.0: Mon Mar  6 21:00:17 PST 2023; root:xnu-8796.101.5~3/RELEASE_X86_64 x86_64
compiler info: clang: 14.0.3 (clang-1403.0.22.14.1)
libc info: no libc information available
$SHELL (typically, interactive shell): /bin/zsh


[Enabled Hooks]

Re: git diff -U0 header off-by-one error when deleting no lines

[Junio C Hamano]

Lukas Tenbrink <lukastenbrink@gmail.com> writes:

> Thank you for filling out a Git bug report!
> Please answer the following questions to help us understand your issue.
>
> What did you do before the bug happened? (Steps to reproduce your issue)
>
> Full reproduction:
>
> echo "A\nB\nC" > test1
> echo "A\nB\nC\nD\nE\nF" > test2

For many folks, the above will place one line each in these files.
For portability, you'd need to do something like

	printf "%s\n" A B C >test1
	printf "%s\n" A B C D E F >test2

> git diff -U0 test1 test2

> The actual output places the new text at line 3:
>
>> git diff -U0 test1 test2
> diff --git a/test1 b/test2
> index b1e6722..cead32e 100644
> --- a/test1
> +++ b/test2
> @@ -3,0 +4,3 @@ C
> +D
> +E
> +F

I think this is very much in line with how "diff -U0" by other
people do it, and any tool like "git am" that need to read "diff"
output need to understand how these output work correctly; "-U0"
output and ",1" that is omitted are things that have confused us
tool writers forever ;-).

With -U0, you'd get

    $ diff -U0 test1 test2
    --- test1       2023-05-03 14:02:27.718960038 -0700
    +++ test2       2023-05-03 14:02:20.094156573 -0700
    @@ -3,0 +4,3 @@
    +D
    +E
    +F

which has "-3,0" that says "there were 3 lines before this hunk that
did not have any line before the change".

This might be unintuitive, but it is specified by POSIX.  

Open

  https://pubs.opengroup.org/onlinepubs/9699919799/

and look for "@@".

    "@@-%s+%s@@", <file1 range>, <file2 range>

    Each <range> field shall be of the form:

    "%1d", <beginning line number>

    or:

    "%1d,1", <beginning line number>

    if the range contains exactly one line, and:

    "%1d,%1d", <beginning line number>, <number of lines>

    otherwise.

    If a range is empty, its beginning line number shall be the
    number of the line just before the range, or 0 if the empty
    range starts the file.

So a patch that adds lines to an empty file would look like

    $ diff -U0 /dev/null test1
    --- /dev/null   2023-04-29 22:24:54.395999895 -0700
    +++ test1       2023-05-03 14:02:27.718960038 -0700
    @@ -0,0 +1,3 @@
    +A
    +B
    +C

just as specified.  Note "-0,0" that says "there was no line before
this hunk before this change".

Re: git diff -U0 header off-by-one error when deleting no lines

[Lukas Tenbrink]

> This might be unintuitive, but it is specified by POSIX.  
> 
> Open
> 
>  https://pubs.opengroup.org/onlinepubs/9699919799/
> 
> and look for "@@".
> 
>    "@@-%s+%s@@", <file1 range>, <file2 range>
> 
>    Each <range> field shall be of the form:
> 
>    "%1d", <beginning line number>
> 
>    or:
> 
>    "%1d,1", <beginning line number>
> 
>    if the range contains exactly one line, and:
> 
>    "%1d,%1d", <beginning line number>, <number of lines>
> 
>    otherwise.
> 
>    If a range is empty, its beginning line number shall be the
>    number of the line just before the range, or 0 if the empty
>    range starts the file.
> 
> So a patch that adds lines to an empty file would look like
> 
>    $ diff -U0 /dev/null test1
>    --- /dev/null   2023-04-29 22:24:54.395999895 -0700
>    +++ test1       2023-05-03 14:02:27.718960038 -0700
>    @@ -0,0 +1,3 @@
>    +A
>    +B
>    +C
> 
> just as specified.  Note "-0,0" that says "there was no line before
> this hunk before this change".


That is indeed fairly unintuitive, but I appreciate the detailed response. I shall adapt my scripts.
Thanks and best wishes,

Lukas