[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#660409: marked as done (Rewrite Britney's installability tester in python)

Your message dated Mon, 02 Dec 2013 21:05:13 +0100
with message-id <529CE7F9.4010700@thykier.net>
and subject line Re: Bug#660409: Rewrite Britney's installability tester in python
has caused the Debian Bug report #660409,
regarding Rewrite Britney's installability tester in python
to be marked as done.

This means that you claim that the problem has been dealt with.
If this is not the case it is now your responsibility to reopen the
Bug report if necessary, and/or fix the problem forthwith.

(NB: If you are a system administrator and have no idea what this
message is talking about, this may indicate a serious mail system
misconfiguration somewhere. Please contact owner@bugs.debian.org
immediately.)


-- 
660409: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=660409
Debian Bug Tracking System
Contact owner@bugs.debian.org with problems
--- Begin Message --- 
Package: release.debian.org
Severity: wishlist
User: release.debian.org@packages.debian.org
Usertags: britney


Hi,

I took a look at replacing Britney's installability tester[1].  In my
follow up mail I will attach my patches[2].  Part of my focus has been
to implement this as a(n almost) compatible replacement for the old
module.

Result-wise it gives identical results to our current implementation
in the test suite.  Speed-wise, I believe it is "fast enough" to be
used.  I have some ideas for further optimization, though these will
most likely imply structural changes to other parts of Britney.

My testing suggests that using Python2.7 will greatly affect the
runtime with this module.  The master branch correctly only see a
similar improvement in one of the live-data samples[3].

Please note that the britney-tests repository is currently not setup
to use the live-data as regression test.  If you want to verify my
results, consider using "test-branch" to pit it against the master
branch.

~Niels

[1] The "dark corner" written in C that no ones ever looks at
if they can avoid it.

[2] I do not trust myself with reportbug to get it right in this
email.  :)

[3] Runtimes the master branch vs my branch.  The "-py2.6" marker means
Python2.6 was used as interpreter.  Same for "-py2.7" and Python2.7.

+------------+--------------+--------------+--------------+--------------+
| live-data  | master-py2.6 | master-py2.7 | branch-py2.6 | branch-py2.7 |
+------------+--------------+--------------+--------------+--------------+
| 2011-12-13 |   201.363s   |   186.369s   |   173.224s   |   135.778s   |
| 2011-12-20 |    97.947s   |    93.213s   |   142.128s   |   111.234s   |
| 2012-01-04 |    74.255s   |    70.503s   |   117.565s   |    95.287s   |
+------------+--------------+--------------+--------------+--------------+

NB: 2011-12-13 triggers an AIEEE in the master branch.

--- End Message ---
--- Begin Message --- 
On 2013-09-18 22:10, Niels Thykier wrote:
> Hi,
> 
> I am back again, with my branch rewritten and rebased.
> 
> The rewritten installability tester uses significantly less memory than
> the current C-module.  In her current state, she tends to exceed 2.5GB
> in the live-data tests.  With these patches she stops at 1.2-1.3GB as
> far as I can trace with random sampling via top(1).
> 
> Review welcome,
> ~Niels
> 
> 

Merged into master on the 27th of November.

~Niels

--- End Message ---
Reply to: