[RFH] plumber's puzzle
From: Junio C Hamano <hidden>
Date: 2016-06-15 22:43:06
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?