[RFH] plumber's puzzle

[Junio C Hamano <junkio@cox.net>]

I know people had trouble fetching from the official git.git
repository this morning; sorry about the gotcha.

Now it has all been hopefully sorted out, here is one thing that
I have been having trouble with lately, and can use some help.

Please try this, after getting the latest 'pu':

	$ git checkout origin/pu
	$ make all test-chmtime
	$ cd t
        $ sh ./t0021-filter.sh
        $ cd trash
        $ rm -f test.t
        $ ../../git-checkout-index -f -q -a
	$ ../../git-diff

This shows the single "diff --git" header line without anything,
to show that the path is not stat-clean, but the contents are
unchanged, which is what is expected.

A bit of background.  test.t has a funny "rot13" filter set for
both smudge and clean.  So test.t file has "a b c d e...", but
the actual blob records "n o p q r...".  Since we teached git
about the filter, git-checkout-index and git-diff do the right
thing.

Now, go back to the toplevel, comment out "close(1)" in
convert.c::filter_buffer() around line 236, and rebuild.

	$ cd ../..
        $ edit convert.c ;# comment out "close(1)"
        $ make
        $ cd t/trash

We get a quite _WRONG_ diff, like this:

        $ ../../git-diff
        diff --git a/test.t b/test.t
        --- a/test.t
        +++ b/test.t
        @@ -1,3 +1,4 @@
         n o p q r s t u v w x y z
         a b c d e f g h i j k l m
         $vqrag$
        +diff --git a/test.t b/test.t

and if you disable pager like this:

	$ PAGER= ../../git-diff

then you get the expected correct result.  Also you get this if
we disable the internal pager:

        $ ../../git-diff | cat
        diff --git a/test.t b/test.t
        --- a/test.t
        +++ b/test.t
        @@ -1,3 +1,4 @@
         n o p q r s t u v w x y z
         a b c d e f g h i j k l m
         $vqrag$
        +diff --git a/test.t b/test.t

I do not want to have a "fix" that I do not understand *why* it
fixes things in the tree; I couldn't figure out why that
close(1) in that codepath matters, although I was the one who
added that line.

Insights?