Debian Bug report logs - #454504
bzr: Large performance regression with rich-root format

Toggle useless messages

Message #5 received at submit@bugs.debian.org (full text, mbox, reply):

From: Roland Mas <lolando@debian.org>

To: Debian Bug Tracking System <submit@bugs.debian.org>

Subject: bzr: Large performance regression with rich-root format

Date: Wed, 05 Dec 2007 20:33:42 +0100

Package: bzr
Version: 1.0~rc1-2
Severity: normal

~roland/debian/bzr-repo/gforge/ holds my main bzr repository, stored
as dirstate-with-subtree with no working trees.  Trying to clone
branches stored in there gives wildly varying performances depending
on the format of the destination repository.  In particular, rich-root
is *slow*:

guest@mirexpress:~/repo/gforge-rich-root$ bzr init-repo --no-trees --format=rich-root .
guest@mirexpress:~/repo/gforge-rich-root$ time bzr branch ~roland/debian/bzr-repo/gforge/debian/sid/ sid
Branched 5068 revision(s).

real	108m52.212s
user	89m24.975s
sys	1m40.582s

[...]

guest@mirexpress:~/repo/gforge-pack$ bzr init-repo --no-trees --format=pack-0.92-subtree .
guest@mirexpress:~/repo/gforge-pack$ time bzr branch ~roland/debian/bzr-repo/gforge/debian/sid/
Branched 5068 revision(s).

real	5m30.865s
user	2m31.997s
sys	0m11.425s

[...]

guest@mirexpress:~/repo/gforge-dirstate$ bzr init-repo --no-trees --format=dirstate-with-subtree .
guest@mirexpress:~/repo/gforge-dirstate$ time bzr branch ~roland/debian/bzr-repo/gforge/upstream-svn/trunk/
Branched 5068 revision(s).

real	4m36.514s
user	1m19.449s
sys	0m13.533s

-- System Information:
Debian Release: lenny/sid
  APT prefers unstable
  APT policy: (500, 'unstable')
Architecture: i386 (i686)

Kernel: Linux 2.6.22-3-k7 (SMP w/1 CPU core)
Locale: LANG=fr_FR.UTF-8, LC_CTYPE=fr_FR.UTF-8 (charmap=UTF-8)
Shell: /bin/sh linked to /bin/bash

Versions of packages bzr depends on:
ii  libc6                         2.7-3      GNU C Library: Shared libraries
ii  python                        2.4.4-6    An interactive high-level object-o
ii  python-celementtree           1.0.5-9    Light-weight toolkit for XML proce
ii  python-central                0.5.15     register and build utility for Pyt

Versions of packages bzr recommends:
ii  python-paramiko               1.6.4-1    make SSH2 connections with python
ii  python-pycurl                 7.16.4-1   Python bindings to libcurl

-- no debconf information

Message #22 received at 454504-close@bugs.debian.org (full text, mbox, reply):

From: Jelmer Vernooij <jelmer@vernstok.nl>

To: 454504-close@bugs.debian.org

Subject: Unable to fix

Date: Wed, 29 Apr 2009 19:55:27 +0200

The data formats you're converting between are significantly different
so bzr can't just copy abstract data, it has to process everything to
convert it.

--pack-0.92 to --rich-root-pack is a lot quicker (you are converting
to --rich-root, which is the rich root equivalent of --knit).

The recent work on using a rich root format by default has made it a
lot less necessary to work with different formats, so that should at
least reduce the likeliness of having multiple formats that bzr has to
convert between.

Alternatively, do you think it would perhaps be useful if bzr warned
when it had to do a slow conversion when fetching?

Message #27 received at 454504@bugs.debian.org (full text, mbox, reply):

From: Roland Mas <lolando@debian.org>

To: jelmer@debian.org

Cc: 454504@bugs.debian.org

Subject: Re: Bug#454504 closed by Jelmer Vernooij <jelmer@vernstok.nl> (Unable to fix)

Date: Wed, 29 Apr 2009 22:11:06 +0200

Jelmer Vernooij:

> The recent work on using a rich root format by default has made it a
> lot less necessary to work with different formats, so that should at
> least reduce the likeliness of having multiple formats that bzr has to
> convert between.

  Fair enough.

> Alternatively, do you think it would perhaps be useful if bzr warned
> when it had to do a slow conversion when fetching?

  Maybe, but if the dichotomy slowly disappears and most of the
repositories are either created or converted to the rich-root world,
then I can understand that there's no real need to go to extra effort
for what's essentially a disappearing problem.  Unless the rich-root
variants won't be made the default for some time yet.

Roland.
-- 
Roland Mas

$ chown -R us:us your_base*

Message #32 received at 454504@bugs.debian.org (full text, mbox, reply):

From: Wouter van Heyst <larstiq@larstiq.dyndns.org>

To: Roland Mas <lolando@debian.org>, 454504@bugs.debian.org

Subject: Re: [Pkg-bazaar-maint] Bug#454504: closed by Jelmer Vernooij <jelmer@vernstok.nl> (Unable to fix)

Date: Wed, 29 Apr 2009 22:27:26 +0200

On Wed, Apr 29, 2009 at 10:11:06PM +0200, Roland Mas wrote:
> Jelmer Vernooij:
> 
> > The recent work on using a rich root format by default has made it a
> > lot less necessary to work with different formats, so that should at
> > least reduce the likeliness of having multiple formats that bzr has to
> > convert between.
> 
>   Fair enough.
> 
> > Alternatively, do you think it would perhaps be useful if bzr warned
> > when it had to do a slow conversion when fetching?
> 
>   Maybe, but if the dichotomy slowly disappears and most of the
> repositories are either created or converted to the rich-root world,
> then I can understand that there's no real need to go to extra effort
> for what's essentially a disappearing problem.  Unless the rich-root
> variants won't be made the default for some time yet.

The bzrtools upstream repo has been converted to a rich-root format just
now, the idea is to see how that pans out, incorporate lessons learned,
upgrade bzr.dev, disseminate.

It won't all happen this week, but I expect things to be rolling by
1.17.

Wouter van Heyst

Message #37 received at 454504@bugs.debian.org (full text, mbox, reply):

From: Jelmer Vernooij <jelmer@debian.org>

To: Roland Mas <lolando@debian.org>

Cc: 454504@bugs.debian.org

Subject: Re: Bug#454504 closed by Jelmer Vernooij <jelmer@vernstok.nl> (Unable to fix)

Date: Wed, 29 Apr 2009 23:27:38 +0200

Roland Mas wrote:
> Jelmer Vernooij:
>   
>> Alternatively, do you think it would perhaps be useful if bzr warned
>> when it had to do a slow conversion when fetching?
>>     
>
>   Maybe, but if the dichotomy slowly disappears and most of the
> repositories are either created or converted to the rich-root world,
> then I can understand that there's no real need to go to extra effort
> for what's essentially a disappearing problem.  Unless the rich-root
> variants won't be made the default for some time yet.
>   
Yeah, the plan from upstream is to have this happen soon. squeeze will
have a rich-root format by default one way or another, if I can help it.

Cheers,

Jelmer

Send a report that this bug log contains spam.