[BUG] Changesets that create empty files produce bogus patches

[Aaron Bentley]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

John A Meinel wrote:
> Well, if you do:
> touch a
> diff -u /dev/null a
> 
> You get no output.

Yes, but I don't think anyone would claim no output is a patch.  Patches
have a particular format, and the empty string doesn't match it.  In
fact, diff produces a different status code when the files are the same,
as well as failing to emit a patch.

>>No patch.  A patch to add an empty file might look like:
>>--- /dev/null
>>+++ mod/sub/sub/emptyfile.txt
>>@@ -0,0 +1,1 @@
>>+
>>\ No newline at end of file
> 
> 
> I don't know that the above is a genuine statement.

If you do echo -n a > nearlyemptyfile.txt and diff it, you'll get this:
- --- /dev/null
+++ mod/sub/sub/nearlyemptyfile.txt
@@ -0,0 +1,1 @@
+ a
\ No newline at end of file

So what I wrote was, I thought, a pretty obvious extrapolation.

> I would say that we
> could easily just realize that an "add" with no content means we should
> create an empty file.

Yes.  If you just didn't invoke note_patch, it would make more sense.

> I think the above patch looks pretty ugly, and doesn't really make it
> clear that it is an empty file. (It could have whitespace)

I certainly have no need to champion bizarre patch kludges like that.
But at least it looks like a patch.  As for whitespace, I'm not sure how
that's different from other patches.  Whitespace differences are rarely
obvious.  Colourizing patches helps.

Aaron
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org

iD8DBQFC6Qz70F+nu1YWqI0RAkBWAJ4iLF2RxqqhJkuyzKL9v63lA9835QCeObXA
rbe99npa3CycRywtlKtNPL4=
=g2tq
-----END PGP SIGNATURE-----