threads / discuss / 3085

gitweb on kernel.org broken

Subject: gitweb on kernel.org broken

## tl;dr

5 messages between Jan 16, 2006 and Jan 16, 2006.

replies: 4people: 4as markdown or json

H. Peter Anvin· Jan 16, 2006, 18:44 UTC · lore

It seems that gitweb throws an error every time. This means that nothing gets cached, and therefore the load on the kernel.org machines is driven into the high double digits:

[Mon Jan 16 10:42:01 2006] [error] [client 82.181.110.11] [Mon Jan 16 10:42:01 2006] index.cgi: Use of uninitialized value in scalar assignment at /home/httpd/html/git/index.cgi line 411. [Mon Jan 16 10:42:03 2006] [error] [client 82.181.110.11] fatal: Not a git repository: '/pub/scm/linux/hotplug/hotplug-ng.git' [Mon Jan 16 10:42:03 2006] [error] [client 82.181.110.11] [Mon Jan 16 10:42:03 2006] index.cgi: Use of uninitialized value in scalar chomp at /home/httpd/html/git/index.cgi line 406. [Mon Jan 16 10:42:03 2006] [error] [client 82.181.110.11] [Mon Jan 16 10:42:03 2006] index.cgi: Use of uninitialized value in pattern match (m//) at /home/httpd/html/git/index.cgi line 407.

This is all very bad...
	-hpa
Kay Sievers· Jan 16, 2006, 18:53 UTC · re: H. Peter Anvin · lore

Re: gitweb on kernel.org broken

On Mon, Jan 16, 2006 at 10:44:35AM -0800, H. Peter Anvin wrote:
Show 17 quoted lines
> It seems that gitweb throws an error every time.  This means that 
> nothing gets cached, and therefore the load on the kernel.org machines 
> is driven into the high double digits:
> 
> [Mon Jan 16 10:42:01 2006] [error] [client 82.181.110.11] [Mon Jan 16 
> 10:42:01 2006] index.cgi: Use of uninitialized value in scalar 
> assignment at /home/httpd/html/git/index.cgi line 411.
> [Mon Jan 16 10:42:03 2006] [error] [client 82.181.110.11] fatal: Not a 
> git repository: '/pub/scm/linux/hotplug/hotplug-ng.git'
> [Mon Jan 16 10:42:03 2006] [error] [client 82.181.110.11] [Mon Jan 16 
> 10:42:03 2006] index.cgi: Use of uninitialized value in scalar chomp at 
> /home/httpd/html/git/index.cgi line 406.
> [Mon Jan 16 10:42:03 2006] [error] [client 82.181.110.11] [Mon Jan 16 
> 10:42:03 2006] index.cgi: Use of uninitialized value in pattern match 
> (m//) at /home/httpd/html/git/index.cgi line 407.
> 
> This is all very bad...
Junio, seems your "textual symref HEAD" change is causing this?
  http://www.kernel.org/git/?p=git/gitweb.git;a=blob;hb=HEAD;f=gitweb.cgi#l399

Thanks, Kay

Kay Sievers· Jan 16, 2006, 19:05 UTC · re: Kay Sievers · lore

Re: gitweb on kernel.org broken

On Mon, Jan 16, 2006 at 07:53:42PM +0100, Kay Sievers wrote:
Show 21 quoted lines
> On Mon, Jan 16, 2006 at 10:44:35AM -0800, H. Peter Anvin wrote:
> > It seems that gitweb throws an error every time.  This means that 
> > nothing gets cached, and therefore the load on the kernel.org machines 
> > is driven into the high double digits:
> > 
> > [Mon Jan 16 10:42:01 2006] [error] [client 82.181.110.11] [Mon Jan 16 
> > 10:42:01 2006] index.cgi: Use of uninitialized value in scalar 
> > assignment at /home/httpd/html/git/index.cgi line 411.
> > [Mon Jan 16 10:42:03 2006] [error] [client 82.181.110.11] fatal: Not a 
> > git repository: '/pub/scm/linux/hotplug/hotplug-ng.git'
> > [Mon Jan 16 10:42:03 2006] [error] [client 82.181.110.11] [Mon Jan 16 
> > 10:42:03 2006] index.cgi: Use of uninitialized value in scalar chomp at 
> > /home/httpd/html/git/index.cgi line 406.
> > [Mon Jan 16 10:42:03 2006] [error] [client 82.181.110.11] [Mon Jan 16 
> > 10:42:03 2006] index.cgi: Use of uninitialized value in pattern match 
> > (m//) at /home/httpd/html/git/index.cgi line 407.
> > 
> > This is all very bad...
> 
> Junio, seems your "textual symref HEAD" change is causing this?
>   http://www.kernel.org/git/?p=git/gitweb.git;a=blob;hb=HEAD;f=gitweb.cgi#l399
I've comment "strict" and "warnings" out, until it's fixed proper.
Kay
Junio C Hamano· Jan 16, 2006, 20:02 UTC · re: Kay Sievers · lore

Re: gitweb on kernel.org broken

Kay Sievers <kay.sievers@vrfy.org> writes:
>> Junio, seems your "textual symref HEAD" change is causing this?

Yeah, I assumed git_read_head() is called only on a repository that was verified to be valid, and the subroutine did not check the case where HEAD does not verify. My bad. The attached patch should solve this.

The error log says hotplug-ng.git is triggering it. I wonder what is so special about that repository. I hope it is transient during mirroring that mirrored HEAD before objects, not a corrupt repository at the mirroring source.

	... goes to kernel.org mirroring source machine ...

Ugh. It is _not_ a valid git repository. HEAD should be either a symlink or a symref pointing somewhere under refs/heads/, but in that directory, it instead is a regular file that contains an object name. There is nothing under refs/ either:

        $ ls -l /pub/scm/linux/hotplug/hotplug-ng.git/HEAD
        -rw-r--r--  2 gregkh users 41 May  6  2005 /pub/s...ug-ng.git/HEAD
	$ cat /pub/scm/linux/hotplug/hotplug-ng.git/HEAD
        b26bec9f620a969fe5238e8666ab698f197df488
	$ ls -l /pub/scm/linux/hotplug/hotplug-ng.git/refs
	total 0
	$ exit

Greg, could you let me know what is going on with this repository, please?

First of all, I am wondering if this repository is still used (maintained); a regular file HEAD that contains an object name would have stopped working on Aug 27 2005 with this commit at the latest, maybe even before that:

    diff-tree 5f5608b... (from 9740d28...
    Author: Linus Torvalds <torvalds@osdl.org>
    Date:   Sat Aug 27 13:54:42 2005 -0700
    [PATCH] Make .git directory validation code test HEAD
    Inspired by a report by Kalle Valo, this changes git-sh-setup-script and
    the "setup_git_directory()" function to test that $GIT_DIR/HEAD is a
    symlink, since a number of core git features depend on that these days.
    ...

The way I expected to see that part of the directory laid out in a valid git repository would be:

	$ cd /pub/scm/linux/hotplug/hotplug-ng.git/
	$ mkdir -p refs/heads
	$ cat HEAD >refs/heads/master
	$ rm -f HEAD
	$ ln -s refs/heads/master HEAD
Anyhow, here is a gitweb patch to skip this.
---
diff --git a/gitweb.cgi b/gitweb.cgi
index 986d7da..f3853e7 100755
--- a/gitweb.cgi
+++ b/gitweb.cgi
@@ -404,9 +404,8 @@ sub git_read_head {
 	if (open my $fd, "-|", "$gitbin/git-rev-parse", "--verify", "HEAD") {
 		my $head = <$fd>;
 		close $fd;
-		chomp $head;
-		if ($head =~ m/^[0-9a-fA-F]{40}$/) {
-			$retval = $head;
+		if (defined $head && $head =~ /^([0-9a-fA-F]{40})$/) {
+			$retval = $1;
 		}
 	}
 	$ENV{'GIT_DIR'} = $oENV;
Greg KH· Jan 16, 2006, 20:28 UTC · re: Junio C Hamano · lore

Re: gitweb on kernel.org broken

On Mon, Jan 16, 2006 at 12:02:05PM -0800, Junio C Hamano wrote:
Show 36 quoted lines
> Kay Sievers <kay.sievers@vrfy.org> writes:
> 
> >> Junio, seems your "textual symref HEAD" change is causing this?
> 
> Yeah, I assumed git_read_head() is called only on a repository
> that was verified to be valid, and the subroutine did not check
> the case where HEAD does not verify.  My bad.  The attached
> patch should solve this.
> 
> The error log says hotplug-ng.git is triggering it.  I wonder what
> is so special about that repository.  I hope it is transient
> during mirroring that mirrored HEAD before objects, not a
> corrupt repository at the mirroring source.
> 
> 	... goes to kernel.org mirroring source machine ...
> 
> Ugh.  It is _not_ a valid git repository.  HEAD should be either a symlink
> or a symref pointing somewhere under refs/heads/, but in that
> directory, it instead is a regular file that contains an object
> name.  There is nothing under refs/ either:
> 
>         $ ls -l /pub/scm/linux/hotplug/hotplug-ng.git/HEAD
>         -rw-r--r--  2 gregkh users 41 May  6  2005 /pub/s...ug-ng.git/HEAD
> 	$ cat /pub/scm/linux/hotplug/hotplug-ng.git/HEAD
>         b26bec9f620a969fe5238e8666ab698f197df488
> 	$ ls -l /pub/scm/linux/hotplug/hotplug-ng.git/refs
> 	total 0
> 	$ exit
> 
> Greg, could you let me know what is going on with this
> repository, please?
> 
> First of all, I am wondering if this repository is still used
> (maintained); a regular file HEAD that contains an object name
> would have stopped working on Aug 27 2005 with this commit at the
> latest, maybe even before that:

That's a very old git tree, and the development tree is pretty much dead too. I'll just go delete the thing.

Sorry about that.
greg k-h

← back to recent threads