git apply does not honor diff.noprefix config setting

hIpPy· May 19, 2018, 23:18 UTC · lore

If I disable mnemonic prefix,

$ git config --global diff.noprefix true

and do a round-trip of format-patch and apply,

$ git format-patch -1 @ $ git apply <some-file.patch>

git apply fails with,

error: git diff header lacks filename information when removing 1 leading pathname component (line 16)

Without 'diff.noprefix' config setting, git apply works. It seems git apply does not honor the diff.noprefix config setting.

Thanks, Rish

Junio C Hamano· May 21, 2018, 02:03 UTC · re: hIpPy · lore

Re: git apply does not honor diff.noprefix config setting

hIpPy <hippy2981@gmail.com> writes:

Show 5 quoted lines

> If I disable mnemonic prefix,
>
> $ git config --global diff.noprefix true
>
> and do a round-trip of format-patch and apply,

Setting diff.noprefix does not disable "mnemonic prefix". It asks "diff" family of commands to use no prefix, not even the normal, non-mnemonic, prefix.

Show 7 quoted lines

> $ git format-patch -1 @
> $ git apply <some-file.patch>
>
> git apply fails with,
>
> error: git diff header lacks filename information when removing 1
> leading pathname component (line 16)

Totally expected.

> Without 'diff.noprefix' config setting, git apply works. It seems git
> apply does not honor the diff.noprefix config setting.

Yes, and because "diff" and "format-patch" are for producers of, and "apply" and "am" are for consumers of a patch, which are likely to be different people, "apply" or "am" should never pay attention to "diff.noprefix".

It is a different issue if we should have

 - format-patch.noprefix, which defaults to the same as
   diff.noprefix, but allows people to configure "format-patch"
   differently from "diff" and "show"

 - am.pvalue, which defaults to 1, but can be set to e.g. 0 to
   accept format-patch output from those who set
   format-patch.noprefix to true

I haven't thought things through, but offhand I do not see why we shouldn't. But I am reasonably firm on that diff.noprefix should never affect anything on the "apply/am" side.