Code review comment for lp:~toshio/bzr/allow-dirty-patches

This looks ok to me. I'm a bit surprised that we stop as soon as we get trailing garbage, but we allow preceding garbage. What happens with a patch like:

=== modified file 'file-1'
--- file-1 2010-01-13 21:15:55 +0000
+++ file-1 2010-02-09 19:19:46 +0000
@@ -54,17 +54,21 @@
 text
+one
-two

A comment about this next change
=== modified file 'file-2'
--- file-1 2010-01-13 21:15:55 +0000
+++ file-1 2010-02-09 19:19:46 +0000
@@ -54,17 +54,21 @@

Are they just considered separate 'patches', and thus the line is treated as bogus header data for the second patch?

How does 'patch' handle interleaved stuff?

Having thought about it more, we may want to be more relaxed overall about what we allow. I guess the original argument was that we wanted to make sure to understand the diff in a merge-proposal, so that we could validate that the diff you preview is the diff that would be applied once merged.