Re: What's cooking in git.git (May 2012, #06; Wed, 23)

Johannes Sixt [off-list ref] writes:

Am 5/24/2012 7:36, schrieb René Scharfe:

quoted

Am 24.05.2012 00:24, schrieb Junio C Hamano:

quoted

* rs/maint-grep-F (2012-05-21) 4 commits
   (merged to 'next' on 2012-05-23 at b24143c)
  + grep: stop leaking line strings with -f
  + grep: support newline separated pattern list
  + grep: factor out do_append_grep_pat()
  + grep: factor out create_grep_pat()

"git grep -F", unlike the case where it reads from a file, did not treat
individual lines in the given pattern argument as separate patterns as it
should.

This is not specific to -F; grep(1) accepts newline-separated pattern
lists with -E etc. as well, as does git grep with the patches above.

Shouldn't we worry that this change breaks existing users?

Consider a script that generates a pattern that sometimes contains NL.
Yes, it would not match anywhere, but what if that is a deliberate choice
of the script writer? With this change, the script would now observe
spurious matches where earlier there were no matches.

While I prefer to be conservative, and the scenario above may be
theoretically possible, I personally find this particular "breakage" of
backward compatibility is very well defensible, as the original behaviour
is clearly buggy and unintuitive.