Once we find the absolute paths for git_dir and work_tree, we can make
git_dir a relative path since we know pwd will be work_tree. This should
save the kernel some time traversing the path to work_tree all the time
if git_dir is inside work_tree.

Signed-off-by: Daniel Barkalow <barkalow@iabervon.org>
---
Not only was work_tree potentially NULL, it was also already a copied 
canonical path. So this simpler patch should be better.

 cache.h |    1 +
 path.c  |   17 +++++++++++++++++
 setup.c |    3 ++-
 3 files changed, 20 insertions(+), 1 deletions(-)

diff --git a/cache.h b/cache.h
index eab1a17..3465c94 100644
--- a/cache.h
+++ b/cache.h
@@ -524,6 +524,7 @@ static inline int is_absolute_path(const char *path)
 	return path[0] == '/';
 }
 const char *make_absolute_path(const char *path);
+const char *make_relative_path(const char *abs, const char *base);
 
 /* Read and unpack a sha1 file into memory, write memory to a sha1 file */
 extern int sha1_object_info(const unsigned char *, unsigned long *);
diff --git a/path.c b/path.c
index b7c24a2..790d8d4 100644
--- a/path.c
+++ b/path.c
@@ -294,6 +294,23 @@ int adjust_shared_perm(const char *path)
 /* We allow "recursive" symbolic links. Only within reason, though. */
 #define MAXDEPTH 5
 
+const char *make_relative_path(const char *abs, const char *base)
+{
+	static char buf[PATH_MAX + 1];
+	int baselen;
+	if (!base)
+		return abs;
+	baselen = strlen(base);
+	if (prefixcmp(abs, base))
+		return abs;
+	if (abs[baselen] == '/')
+		baselen++;
+	else if (base[baselen - 1] != '/')
+		return abs;
+	strcpy(buf, abs + baselen);
+	return buf;
+}
+
 const char *make_absolute_path(const char *path)
 {
 	static char bufs[2][PATH_MAX + 1], *buf = bufs[0], *next_buf = bufs[1];
diff --git a/setup.c b/setup.c
index d630e37..1643ee4 100644
--- a/setup.c
+++ b/setup.c
@@ -292,7 +292,8 @@ void setup_work_tree(void)
 	work_tree = get_git_work_tree();
 	git_dir = get_git_dir();
 	if ...
[ message continues ]

Hi,



All in all I am pretty surprised how easy it was.  I tried yesterday, for 
half an hour, to come up with something sensible, and failed.

Thanks,
Dscho

--

[ message continues ]

I'm not clear on the semantics of !get_git_work_tree(); is a non-absolute 

I was sure you'd come up with just this solution, because you'd just 
recently explained that make_absolute_path() means you can find when one 
path is in another path with a simple string compare. And, since we know 
what pwd is going to be...

	-Daniel
*This .sig left intentionally blank*
--

[ message continues ]

Hi,


My reading was: if there is no work_tree, then a relative git_dir is just 
fine, since we are quite unlikely to jump around in the file system.

And your implementation of make_relative_path() is nice enough to a 
(work_tree ==) base == NULL, but would return the absolute path in that 
case.

Haven't had time to test anything, though.

Ciao,
Dscho
--

[ message continues ]

This special case may help the specific caller you have below, but doesn't
it make the function do more than it advertises with its name?

Other than that, I think the change is Ok, but as a "performance tweak",
--

[ message continues ]

I don't think so; the best path relative to nothing is the absolute path. 
The idea is to return the easiest equivalent path if your pwd is known to 
be "base" and you give it an absolute path. If you don't know what your 
pwd is, the easiest equivalent path is the absolute path with no symlinks. 
Similarly, you get an absolute path if the path you give it isn't inside 

I was hoping Linus would provide some, since he'd noticed the slowness in 
the first place. I'm not sure I have the RAM to have the kernel spend 
non-trivial time looking up path elements in an all-cached tree. I'll try 
to replicate Linus's test case when I get a chance if he hasn't said 
anything.

	-Daniel
*This .sig left intentionally blank*
--

[ message continues ]

Daniel's patch didn't apply for me as-is, so I recreated it with some 
differences (appended), and here are the numbers from ten runs each.

There is some IO for me - probably due to more-or-less random flushing of 
the journal - so the variation is bigger than I'd like, but whatever:

	Before:
		real    0m8.135s
		real    0m7.933s
		real    0m8.080s
		real    0m7.954s
		real    0m7.949s
		real    0m8.112s
		real    0m7.934s
		real    0m8.059s
		real    0m7.979s
		real    0m8.038s


	After:
		real    0m7.685s
		real    0m7.968s
		real    0m7.703s
		real    0m7.850s
		real    0m7.995s
		real    0m7.817s
		real    0m7.963s
		real    0m7.955s
		real    0m7.848s
		real    0m7.969s

Now, going by "best of ten" (on the assumption that the longer numbers 
are all due to IO), I'm saying a 7.933s -> 7.685s reduction, and it does 
seem to be outside of the noise (ie the "after" case never broke 8s, while 
the "before" case did so half the time).

So looks like about 3% to me. 

Doing it for a slightly smaller test-case (just the "arch" subdirectory) 
gets more stable numbers probably due to not filling the journal with 
metadata updates, so we have:

	Before:
		real    0m1.633s
		real    0m1.633s
		real    0m1.633s
		real    0m1.632s
		real    0m1.632s
		real    0m1.630s
		real    0m1.634s
		real    0m1.631s
		real    0m1.632s
		real    0m1.632s

	After:
		real    0m1.610s
		real    0m1.609s
		real    0m1.610s
		real    0m1.608s
		real    0m1.607s
		real    0m1.610s
		real    0m1.609s
		real    0m1.611s
		real    0m1.608s
		real    0m1.611s

where I'ld just take the averages and say 1.632 vs 1.610, which is just 
over 1% peformance improvement.

So it's not in the noise, but it's not as big as I initially thought and 
measured.

(That said, it obviously depends on how deep the working directory path is 
too, and whether it is behind NFS or something else that might need to 
cause more work to look up).

Modified/updated patch ...
[ message continues ]