threads / patch / 32418

patchoneway_merge(): only lstat() when told to update worktree

Subject: [PATCH] oneway_merge(): only lstat() when told to update worktree

## tl;dr

3 messages between Dec 20, 2012 and Dec 20, 2012. Diffs are folded; open one to read it.

replies: 2people: 2as markdown or json

Martin von Zweigbergk· Dec 20, 2012, 17:37 UTC · lore

Although the subject line of 613f027 (read-tree -u one-way merge fix to check out locally modified paths., 2006-05-15) mentions "read-tree -u", it did not seem to check whether -u was in effect. Not checking whether -u is in effect makes e.g. "read-tree --reset" lstat() the worktree, even though the worktree stat should not matter for that operation.

This speeds up e.g. "git reset" a little on the linux-2.6 repo (best of five, warm cache):

        Before      After
real    0m0.288s    0m0.233s
user    0m0.190s    0m0.150s
sys     0m0.090s    0m0.080s
---

I am very unfamiliar with this part of git, so my attempt at a motivation may be totally off.

I have another twenty-or-so patches to reset.c coming up that take the timings down to around 90 ms, but this patch was quite unrelated to that. Those other patches actually make this patch pointless for "git reset" (it takes another path), but I hope this is still a good change for other operations that use oneway_merge.

 unpack-trees.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
Show changes to unpack-trees.c +1 −1
diff --git a/unpack-trees.c b/unpack-trees.c
index 6d96366..61acc5e 100644
--- a/unpack-trees.c
+++ b/unpack-trees.c
@@ -1834,7 +1834,7 @@ int oneway_merge(struct cache_entry **src, struct unpack_trees_options *o)
 
 	if (old && same(old, a)) {
 		int update = 0;
-		if (o->reset && !ce_uptodate(old) && !ce_skip_worktree(old)) {
+		if (o->reset && o->update && !ce_uptodate(old) && !ce_skip_worktree(old)) {
 			struct stat st;
 			if (lstat(old->name, &st) ||
 			    ie_match_stat(o->src_index, old, &st, CE_MATCH_IGNORE_VALID|CE_MATCH_IGNORE_SKIP_WORKTREE))

-- 
1.8.0.1.240.ge8a1f5a
Junio C Hamano· Dec 20, 2012, 20:02 UTC · re: Martin von Zweigbergk · lore

Re: [PATCH] oneway_merge(): only lstat() when told to update worktree

Martin von Zweigbergk <martinvonz@gmail.com> writes:
Show 15 quoted lines
> Although the subject line of 613f027 (read-tree -u one-way merge fix
> to check out locally modified paths., 2006-05-15) mentions "read-tree
> -u", it did not seem to check whether -u was in effect. Not checking
> whether -u is in effect makes e.g. "read-tree --reset" lstat() the
> worktree, even though the worktree stat should not matter for that
> operation.
>
> This speeds up e.g. "git reset" a little on the linux-2.6 repo (best
> of five, warm cache):
>
>         Before      After
> real    0m0.288s    0m0.233s
> user    0m0.190s    0m0.150s
> sys     0m0.090s    0m0.080s
> ---
Sign-off?

I briefly discussed this with Martin in person and came to the same conclusion. To me this looks like an obvious performance fix, but an independent code audit catches our mistakes is of course welcomed.

Thanks.
Show 25 quoted lines
> I am very unfamiliar with this part of git, so my attempt at a
> motivation may be totally off.
>
> I have another twenty-or-so patches to reset.c coming up that take the
> timings down to around 90 ms, but this patch was quite unrelated to
> that. Those other patches actually make this patch pointless for "git
> reset" (it takes another path), but I hope this is still a good change
> for other operations that use oneway_merge.
>
>  unpack-trees.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/unpack-trees.c b/unpack-trees.c
> index 6d96366..61acc5e 100644
> --- a/unpack-trees.c
> +++ b/unpack-trees.c
> @@ -1834,7 +1834,7 @@ int oneway_merge(struct cache_entry **src, struct unpack_trees_options *o)
>  
>  	if (old && same(old, a)) {
>  		int update = 0;
> -		if (o->reset && !ce_uptodate(old) && !ce_skip_worktree(old)) {
> +		if (o->reset && o->update && !ce_uptodate(old) && !ce_skip_worktree(old)) {
>  			struct stat st;
>  			if (lstat(old->name, &st) ||
>  			    ie_match_stat(o->src_index, old, &st, CE_MATCH_IGNORE_VALID|CE_MATCH_IGNORE_SKIP_WORKTREE))
Martin von Zweigbergk· Dec 20, 2012, 21:03 UTC · re: Junio C Hamano · lore

[PATCH v2] oneway_merge(): only lstat() when told to update worktree

Although the subject line of 613f027 (read-tree -u one-way merge fix to check out locally modified paths., 2006-05-15) mentions "read-tree -u", it did not seem to check whether -u was in effect. Not checking whether -u is in effect makes e.g. "read-tree --reset" lstat() the worktree, even though the worktree stat should not matter for that operation.

This speeds up e.g. "git reset" a little on the linux-2.6 repo (best of five, warm cache):

        Before      After
real    0m0.288s    0m0.233s
user    0m0.190s    0m0.150s
sys     0m0.090s    0m0.080s
Signed-off-by: Martin von Zweigbergk <martinvonz@gmail.com>
---
 unpack-trees.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
Show changes to unpack-trees.c +1 −1
diff --git a/unpack-trees.c b/unpack-trees.c
index 6d96366..61acc5e 100644
--- a/unpack-trees.c
+++ b/unpack-trees.c
@@ -1834,7 +1834,7 @@ int oneway_merge(struct cache_entry **src, struct unpack_trees_options *o)
 
 	if (old && same(old, a)) {
 		int update = 0;
-		if (o->reset && !ce_uptodate(old) && !ce_skip_worktree(old)) {
+		if (o->reset && o->update && !ce_uptodate(old) && !ce_skip_worktree(old)) {
 			struct stat st;
 			if (lstat(old->name, &st) ||
 			    ie_match_stat(o->src_index, old, &st, CE_MATCH_IGNORE_VALID|CE_MATCH_IGNORE_SKIP_WORKTREE))

-- 
1.8.0.1.240.ge8a1f5a

← back to recent threads