Inkscape: A Vector Drawing Tool

bzr branch lp:inkscape is very slow and emits an errors and leaves a broken local repository after exit

Reported by Shlomi Fish on 2009-12-19
34
This bug affects 7 people
Affects Status Importance Assigned to Milestone
Inkscape
Medium
Unassigned

Bug Description

Here is a transcript of a shell session with bzr trying to check out lp:inkscape (on my Mandriva Cooker system):

<<<<<<<<<<<<<<<<<<<<<<<<<<
$ time bzr branch lp:inkscape
You have not informed bzr of your Launchpad ID, and you must do this to
write to Launchpad or access private data. See "bzr help launchpad-login".
Format <RepositoryFormatKnit4> for http://bazaar.launchpad.net/~inkscape.dev/inkscape/trunk/.bzr/ is deprecated - please use 'bzr upgrade' to get better performance
[#########| ] 789KB 134KB/s | Fetching revisions:Inserting strea[#########/ ] 790KB 0KB/s | Fetching revisions:Inserting strea[#########- ] 1302KB 133KB/s | Fetching revisions:Inserting strea[#########\ ] 1600KB 126KB/s | Fetching revisions:Inserting strea[#########| ] 1601KB 0KB/s | Fetching revisions:Inserting strea[#########/ ] 1705KB 64KB/s | Fetching revisions:Inserting strea[#########- ] 1705KB 64KB/s | Fetching revisions:Inserting strea[#########\ ] 1893KB 180KB/s | Fetching revisions:Inserting strea[#########|

[SNIPPED]

2KB/s | Fetching revisions:Inserting strea[###################|] 13515KB 1KB/s | Fetching revisions:Inserting strea[###################/] 13518KB 3KB/s | Fetching revisions:Inserting strea[###################-] 13519KB 2KB/s | Fetching revisions:Inserting strea[###################\] 13521KB 1KB/s | Fetching revisions:Inserting strea[###################|] 13526KB 7KB/s | Fetching revisions:Inserting strea[###################/] 13527KB 2KB/s | Fetching revisions:Inserting strea bzr: ERROR: Connection closed: Connection lost while sending request.
Command exited with non-zero status 3
12.97user 0.63system 3:57.23elapsed 5%CPU (0avgtext+0avgdata 189888maxresident)k
176inputs+96outputs (0major+34948minor)pagefaults 0swaps
shlomi:/mnt/old-mandriva/Download/unpack/inkscape/bzr$ ls
inkscape
shlomi:/mnt/old-mandriva/Download/unpack/inkscape/bzr$ cd inkscape/
shlomi:/mnt/old-mandriva/Download/unpack/inkscape/bzr/inkscape$ time bzr pull
bzr: ERROR: Not a branch: "/mnt/old-mandriva/Download/unpack/inkscape/bzr/inkscape/.bzr/branch/".
Command exited with non-zero status 3
1.16user 0.06system 0:01.26elapsed 96%CPU (0avgtext+0avgdata 58192maxresident)k
0inputs+8outputs (0major+4001minor)pagefaults 0swaps
>>>>>>>>>>>>>>>>>>>>>>

So it took bzr checkout lp:inkscape 3 minutes to finish on an error, and it then left the local branch/repository/etc. in an intermediate state. From my impression, checking out large bzr repositories, especially from Launchpad is very slow, and bzr in general is extremely slow. Until now, I used to like to try out various snapshots of the inkscape development using svn, but I fear that having to do the same with bzr will be close to impossible. Not to mention that in bzr even simple commands such as "bzr diff" are excruciatingly slow.

I heavily question the decision to switch to bzr of all version control systems due to its incredible slowness and as a result lack of usability.

jazzynico (jazzynico) wrote :

I confirm bzr *was* very slow, but never encountered that kind of error.
Could you please confirm that the errors are fixed and that bzr is a bit faster now that we've updated the trunk repository and branch formats?

Changed in inkscape:
importance: Undecided → Medium
status: New → Confirmed
Shlomi Fish (shlomif-gmail) wrote :

I'm trying now again on my Mandriva Cooker system (though a similar problem happened also on a Debian VirtualBox VM). I'm running it overnight under "time". It now seems that it's downloading at 1 kB/s or 2 kB/s or so.

I've written a blog post about my frustrations with bzr's performance here:

http://community.livejournal.com/shlomif_tech/41922.html

I admit it's a bit flamatory and other people said that it works better for them on more underpowered machines and on Windows. Nevertheless, mercurial, git and even subversion are very fast here. Someone who is a bzr fan said "I'm doing it wrong." but I'm out of ideas, because bzr on a brand new installation of Debian Testing on a new partition is still very slow.

Regards,

-- Shlomi Fish

Shlomi Fish (shlomif-gmail) wrote :

OK, here are the results:

<quote>
shlomif:~/TEMP/inkscape-checkout$ time bzr branch lp:inkscape
You have not informed bzr of your Launchpad ID, and you must do this to
write to Launchpad or access private data. See "bzr help launchpad-login".
Branched 10077 revision(s).

real 82m42.808s
user 5m15.830s
sys 0m25.767s
</quote>

So it takes a lot of time, but now finishes to the end.

~suv (suv-lp) wrote :

Is this still an issue with current trunk and latest version of bzr?

(Note: 'bzr checkout --lightweight lp:inkscape' instead of 'bzr branch lp:inkscape' reduces the amount of data downloaded. It does slow down however bzr operations later on which need to access the revision history (requires online access) - not really recommended if you work on a feature branch, but ok if you just need the sources to compile latest trunk)

Changed in inkscape:
status: Confirmed → Incomplete
Shlomi Fish (shlomif-gmail) wrote :

@suv: this is no longer an issue for me:

shlomif@telaviv1:~/Download/unpack/graphics/TO_DELETE--inkscape-bzr-branch$ time bzr branch lp:inkscape
Branched 12979 revisions.

real 4m56.140s
user 1m40.436s
sys 0m1.876s

And the inkscape clone seems OK. It may still be an issue for some other people.

~suv (suv-lp) wrote :

Thx for the feedback (much appreciated). Closing this report for now - feel free to reopen if you disagree, or in case you encounter the same issue again and think that it might be related to specific configuration options of Inkscape's bzr repository. Discussions about switching DVCS should be moved to the developers' mailing list.

Changed in inkscape:
status: Incomplete → Invalid
To post a comment you must log in.
This report contains Public information  Edit
Everyone can see this information.

Other bug subscribers