[PATCH 0/4] Repository format version check series.

[PATCH 0/4] Repository format version check series.

[Junio C Hamano]

This series is what I've been cooking for the past several days,
partly based on patch from Martin Atukunda but with hopefully
smaller impact.

[PATCH 1/4] Repository format version check.
[PATCH 2/4] Check repository format version in enter_repo().
[PATCH 3/4] init-db: check template and repository format.
[PATCH 4/4] setup_git_directory(): check repository format version.

The first in the series adds the core.repositoryformatversion
configuration item.  The current repository format version is
defined as version 0, and is the same as not having that in the
configuration file.

The second one adds a repository format version check to
enter_repo().  This function is used to verify the path given by
the user, either from the command line or over the network, is a
valid git repository by daemon, upload-pack (which runs on the
other end when you run fetch-pack or clone-pack), and
receive-pack (which runs on the other end when you run
send-pack).

The third one makes sure we do not reinitialize a repository
whose format version we do not understand, do not copy
template files from unmatching version into a new repository,
and record the repository format version in a newly created
repository.

The last bit in the series adds the same check code to
setup_git_directory().  Along with the earlier change to
git-sh-setup that uses git-var to verify the repository, this
gives the shell scripts the same repository format version
check.

I'll push this out to the "master" sometime over the weekend,
and hopefully merge it into "maint" and cut the next 1.0rc on
Wednesday.  Let's hope that will be the last 1.0rc.

What's in master and pu (aka when will 1.0rc4 be out)

[Junio C Hamano]

Junio C Hamano <junkio@cox.net> writes:

> This series is what I've been cooking for the past several days,
> partly based on patch from Martin Atukunda but with hopefully
> smaller impact.
>
> [PATCH 1/4] Repository format version check.
> [PATCH 2/4] Check repository format version in enter_repo().
> [PATCH 3/4] init-db: check template and repository format.
> [PATCH 4/4] setup_git_directory(): check repository format version.

And they are now in the master branch, along with a couple of
bugfixes I received in the last couple of days.  I've given them
some testing and believe they are in good enough shape for 1.0
futureproofing.  Bug reports with reproduction recipe and/or
patch are most welcomed.

Cooking in the proposed updates branch is the 8-series "work
from subdirectory" patch I posted yesterday, with some polishing
based on suggestion from Linus. The ls-tree rework is not
included at this point, neither is the git-sh-setup, both of
which unfortunately have wider impact than I feel comfortable to
swallow in one day.

I am hoping to base the next "maint" branch update, probably on
Wednesday, on what is in "pu" tonight, with safer updates I will
receive from the list (Documentation fixes and updates, and
archimport updates from Eric and Martin are the candidates).
That will be tagged as 1.0rc4 and hopefully be the last 1.0rc,
which means no more major feature/semantics changes after that
until 1.0 --- famous last words.

So I anticipate that many of the barebone Porcelain commands
that insist on running from the toplevel will ship the way they
are in 1.0.  If you run "git grep git-sh-setup 'git*sh'", you
will notice that most of them are whole-tree operations anyway,
so not much is lost [*1*].

[Footnote]

*1* Bisecting only in a subdirectory might be an interesting
thing to do, hunting down a bug in a subsystem, but usually a
subsystem has at least two relevant subdirectories (source and
include file trees) ;-).

I suspect that it would be trivial to convert whole-repository
operations such as count-objects, prune, and repack to work from
subdirectory, although the value of that is dubious.

Re: What's in master and pu (aka when will 1.0rc4 be out)

[Timo Hirvonen]

On Sun, 27 Nov 2005 02:56:21 -0800
Junio C Hamano <junkio@cox.net> wrote:

> *1* Bisecting only in a subdirectory might be an interesting
> thing to do, hunting down a bug in a subsystem, but usually a
> subsystem has at least two relevant subdirectories (source and
> include file trees) ;-).

Hmm.. would something like 

    git-bisect --include drivers/char/drm/ --include include/ ...

be easy to implement?  That command would limit bisecting to only the
commits that have changed files under those two directories.

-- 
http://onion.dynserv.net/~timo/

Re: What's in master and pu (aka when will 1.0rc4 be out)

[Linus Torvalds]

On Sun, 27 Nov 2005, Timo Hirvonen wrote:
> 
> Hmm.. would something like 
> 
>     git-bisect --include drivers/char/drm/ --include include/ ...
> 
> be easy to implement?  That command would limit bisecting to only the
> commits that have changed files under those two directories.

Heh.

It was surprisingly easy to do.

The syntax is

	git bisect start <pathspec>

followed by all the normal "git bisect good/bad" stuff.

Almost totally untested, and I guarantee that if your pathnames have 
spaces in them (or your GIT_DIR has spaces in it) this won't work. I don't 
know how to fix that, my shell programming isn't good enough.

This involves small changes to make "git-rev-list --bisect" work in the 
presense of a pathspec limiter, and then truly trivial (and that's the 
broken part) changes to make "git bisect" save away and use the pathspec.

I tried one bisection, and a "git bisect visualize", and it all looked 
correct. But hey, don't be surprised if it has problems.

		Linus

---
diff --git a/git-bisect.sh b/git-bisect.sh
index d92993b..2455f00 100755
--- a/git-bisect.sh
+++ b/git-bisect.sh
@@ -33,7 +33,6 @@ bisect_autostart() {
 }
 
 bisect_start() {
-        case "$#" in 0) ;; *) usage ;; esac
 	#
 	# Verify HEAD. If we were bisecting before this, reset to the
 	# top-of-line master first!
@@ -57,7 +56,8 @@ bisect_start() {
 	rm -f "$GIT_DIR/refs/heads/bisect"
 	rm -rf "$GIT_DIR/refs/bisect/"
 	mkdir "$GIT_DIR/refs/bisect"
-	echo "git-bisect start" >"$GIT_DIR/BISECT_LOG"
+	echo "git-bisect start $@" >"$GIT_DIR/BISECT_LOG"
+	echo "$@" > "$GIT_DIR/BISECT_NAMES"
 }
 
 bisect_bad() {
@@ -121,7 +121,7 @@ bisect_next() {
 	bad=$(git-rev-parse --verify refs/bisect/bad) &&
 	good=$(git-rev-parse --sq --revs-only --not \
 		$(cd "$GIT_DIR" && ls refs/bisect/good-*)) &&
-	rev=$(eval "git-rev-list --bisect $good $bad") || exit
+	rev=$(eval "git-rev-list --bisect $good $bad -- $(cat $GIT_DIR/BISECT_NAMES)") || exit
 	if [ -z "$rev" ]; then
 	    echo "$bad was both good and bad"
 	    exit 1
@@ -131,7 +131,7 @@ bisect_next() {
 	    git-diff-tree --pretty $rev
 	    exit 0
 	fi
-	nr=$(eval "git-rev-list $rev $good" | wc -l) || exit
+	nr=$(eval "git-rev-list $rev $good -- $(cat $GIT_DIR/BISECT_NAMES)" | wc -l) || exit
 	echo "Bisecting: $nr revisions left to test after this"
 	echo "$rev" > "$GIT_DIR/refs/heads/new-bisect"
 	git checkout new-bisect || exit
@@ -142,7 +142,7 @@ bisect_next() {
 
 bisect_visualize() {
 	bisect_next_check fail
-	gitk bisect/bad --not `cd "$GIT_DIR/refs" && echo bisect/good-*`
+	gitk bisect/bad --not `cd "$GIT_DIR/refs" && echo bisect/good-*` -- $(cat $GIT_DIR/BISECT_NAMES)
 }
 
 bisect_reset() {
diff --git a/rev-list.c b/rev-list.c
index e17f928..8020d97 100644
--- a/rev-list.c
+++ b/rev-list.c
@@ -350,7 +350,8 @@ static int count_distance(struct commit_
 
 		if (commit->object.flags & (UNINTERESTING | COUNTED))
 			break;
-		nr++;
+		if (!paths || (commit->object.flags & TREECHANGE))
+			nr++;
 		commit->object.flags |= COUNTED;
 		p = commit->parents;
 		entry = p;
@@ -362,6 +363,7 @@ static int count_distance(struct commit_
 			}
 		}
 	}
+
 	return nr;
 }
 
@@ -382,15 +384,20 @@ static struct commit_list *find_bisectio
 	nr = 0;
 	p = list;
 	while (p) {
-		nr++;
+		if (!paths || (p->item->object.flags & TREECHANGE))
+			nr++;
 		p = p->next;
 	}
 	closest = 0;
 	best = list;
 
-	p = list;
-	while (p) {
-		int distance = count_distance(p);
+	for (p = list; p; p = p->next) {
+		int distance;
+
+		if (paths && !(p->item->object.flags & TREECHANGE))
+			continue;
+
+		distance = count_distance(p);
 		clear_distance(list);
 		if (nr - distance < distance)
 			distance = nr - distance;
@@ -398,7 +405,6 @@ static struct commit_list *find_bisectio
 			best = p;
 			closest = distance;
 		}
-		p = p->next;
 	}
 	if (best)
 		best->next = NULL;

Re: What's in master and pu (aka when will 1.0rc4 be out)

[Junio C Hamano]

Linus Torvalds <torvalds@osdl.org> writes:

> The syntax is
>
> 	git bisect start <pathspec>
>
> followed by all the normal "git bisect good/bad" stuff.

I love how I can just say "this might be interesting" and 
find your message in my mailbox next morning [*1*] ;-).

This would be useful to cut down the number of bisection needed
to reach the culprit if the initial guess is correct as to which
part of the tree is involved.

[Reference]
*1* http://marc.theaimsgroup.com/?l=git&m=111669620826189

[PATCH] bisect: quote pathnames for eval safety.

[Junio C Hamano]

Linus Torvalds <torvalds@osdl.org> writes:

> Almost totally untested, and I guarantee that if your pathnames have 
> spaces in them (or your GIT_DIR has spaces in it) this won't work. I don't 
> know how to fix that, my shell programming isn't good enough.

Like this?  I've placed the bisect patch with this in the
proposed updates.

---

 git-bisect.sh |   16 ++++++++++++----
 1 files changed, 12 insertions(+), 4 deletions(-)

applies-to: 7da44ab45773e0a424c1e08c838c1f6deeb6ceb2
82c8f5bed2ae61f1f368cfca84f6b0194b1ad208
diff --git a/git-bisect.sh b/git-bisect.sh
index 2455f00..d4196f9 100755
--- a/git-bisect.sh
+++ b/git-bisect.sh
@@ -1,9 +1,16 @@
 #!/bin/sh
 . git-sh-setup
 
+sq() {
+    for x
+    do
+	echo "$x" | sed -e "s/'/'\\\\''/g" -e "s/^/'/" -e "s/\$/'/"
+    done
+}
+
 usage() {
     echo >&2 'usage: git bisect [start|bad|good|next|reset|visualize]
-git bisect start		reset bisect state and start bisection.
+git bisect start [<pathspec>]	reset bisect state and start bisection.
 git bisect bad [<rev>]		mark <rev> a known-bad revision.
 git bisect good [<rev>...]	mark <rev>... known-good revisions.
 git bisect next			find next bisection to test and check it out.
@@ -56,8 +63,8 @@ bisect_start() {
 	rm -f "$GIT_DIR/refs/heads/bisect"
 	rm -rf "$GIT_DIR/refs/bisect/"
 	mkdir "$GIT_DIR/refs/bisect"
-	echo "git-bisect start $@" >"$GIT_DIR/BISECT_LOG"
-	echo "$@" > "$GIT_DIR/BISECT_NAMES"
+	sq git-bisect start "$@" >"$GIT_DIR/BISECT_LOG"
+	sq "$@" >"$GIT_DIR/BISECT_NAMES"
 }
 
 bisect_bad() {
@@ -142,7 +149,8 @@ bisect_next() {
 
 bisect_visualize() {
 	bisect_next_check fail
-	gitk bisect/bad --not `cd "$GIT_DIR/refs" && echo bisect/good-*` -- $(cat $GIT_DIR/BISECT_NAMES)
+	not=`cd "$GIT_DIR/refs" && echo bisect/good-*`
+	eval gitk bisect/bad --not $not -- $(cat "$GIT_DIR/BISECT_NAMES")
 }
 
 bisect_reset() {
---
0.99.9.GIT

Re: [PATCH] bisect: quote pathnames for eval safety.

[Junio C Hamano]

Junio C Hamano <junkio@cox.net> writes:

> Linus Torvalds <torvalds@osdl.org> writes:
>
>> Almost totally untested, and I guarantee that if your pathnames have 
>> spaces in them (or your GIT_DIR has spaces in it) this won't work. I don't 
>> know how to fix that, my shell programming isn't good enough.
>
> Like this?  I've placed the bisect patch with this in the
> proposed updates.

Not like that ;-).  Fixed one is in pu.

Re: [PATCH 0/4] Repository format version check series.

[Martin Atukunda]

On Fri, Nov 25, 2005 at 05:15:02PM -0800, Junio C Hamano wrote:
> This series is what I've been cooking for the past several days,
> partly based on patch from Martin Atukunda but with hopefully
> smaller impact.
> 
> [PATCH 1/4] Repository format version check.
> [PATCH 2/4] Check repository format version in enter_repo().
> [PATCH 3/4] init-db: check template and repository format.
> [PATCH 4/4] setup_git_directory(): check repository format version.

This is just to confirm that this patch series passed the tests on my
end. Good work.

- Martin -

-- 
Due to a shortage of devoted followers, the production of great leaders has been discontinued.