Debian Bug report logs - #709460
[Britney] binNMUs plus source take-over issue

Toggle useless messages

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

From: Niels Thykier <niels@thykier.net>

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

Subject: [Britney] binNMUs plus source take-over issue

Date: Thu, 23 May 2013 15:06:34 +0200

Package: release.debian.org
Severity: normal
User: release.debian.org@packages.debian.org
Usertags: britney

The story is something like this (Kudos to Adam for debugging this):

  * llvm-3.2 used to produce llvm-3.2-doc/3.2-5 (arch:all)
    - source is in testing and sid
  * llvm-toolchain has taken over llvm-3.2-doc as 1:3.2repack-6
    - the binary is still arch:all
    - source is only in sid
    - source is currently not built on all the architectures (that
      previously built llvm-3.2)
  * llvm-3.2 was binNMU'ed and the binNMU migrates during the run.


The problem here, is that llvm-3.2-doc/1:3.2repack-6 ends up in the
"testing" without its source which causes Britney to crash (when doing
a binary -> source lookup later).


Additional information (copy-waste from IRC):

 src:llvm-3.2 has been superseded by src:llvm-toolchain, on a couple
 of arches. they both build arch:all packages with the same names, so
 dak puts both in the packages files and we throw away the
 lower-versioned one

 then some llvm-3.2 binNMUs (form the old source) migrate during the
 run, and the binary list she copies over includes the new
 arch:all. which she then tries to map back to a source package.

Dak says/said:
 llvm-3.2-doc |         3.2-5 |    testing | all
 llvm-3.2-doc |         3.2-5 |   unstable | all
 llvm-3.2-doc | 1:3.2repack-6 |   unstable | all

~Niels

Message #10 received at 709460@bugs.debian.org (full text, mbox, reply):

From: Niels Thykier <niels@thykier.net>

To: 709460@bugs.debian.org

Subject: Re: Bug#709460: [Britney] binNMUs plus source take-over issue

Date: Thu, 11 Jul 2013 10:05:47 +0200

Control: tags -1 confirmed

On 2013-05-23 15:06, Niels Thykier wrote:
> [...]
> 
> The story is something like this (Kudos to Adam for debugging this):
> 
>   * llvm-3.2 used to produce llvm-3.2-doc/3.2-5 (arch:all)
>     - source is in testing and sid
>   * llvm-toolchain has taken over llvm-3.2-doc as 1:3.2repack-6
>     - the binary is still arch:all
>     - source is only in sid
>     - source is currently not built on all the architectures (that
>       previously built llvm-3.2)
>   * llvm-3.2 was binNMU'ed and the binNMU migrates during the run.
> 
> 
> The problem here, is that llvm-3.2-doc/1:3.2repack-6 ends up in the
> "testing" without its source which causes Britney to crash (when doing
> a binary -> source lookup later).
> 
> 
> [...]
> 
> ~Niels
> 
> 

We now have a test case for this bug[1].

~Niels

[1]
http://anonscm.debian.org/gitweb/?p=collab-maint/britney2-tests.git;a=commitdiff;h=bfa55f0b307711786332d9f3eba6ab14073e2b8e

Message #17 received at 709460@bugs.debian.org (full text, mbox, reply):

From: Anthony Towns <aj@erisian.com.au>

To: 709460@bugs.debian.org

Subject: Re: Bug#709460: [Britney] binNMUs plus source take-over issue

Date: Sun, 26 Apr 2015 09:12:39 +1000

[Message part 1 (text/plain, inline)]

Hi,

Attached is a possible patch for bug#709460. Works against both normal and
live-data test cases by the looks.

N: Using Time::HiRes to calculate run times
Running live-2011-12-13... ok (258.376s)
Running live-2011-12-20... ok (251.626s)
Running live-2012-01-04... ok (193.008s)
Running live-2012-05-09... done (159.661s)

The most notable change is that I pulled the _compute_groups function out
as the first thing that doop_source does, and then reuse its output later
rather than working out which binary packages are valid independently.
(I've left in some commented print statements that add some output as to
what binaries doop_source is actually looking at which seemed helpful)

Cheers,
aj

-- 
Anthony Towns <aj@erisian.com.au>

[Message part 2 (text/html, inline)]

[bug-709460.diff (text/plain, attachment)]

Message #22 received at 709460-done@bugs.debian.org (full text, mbox, reply):

From: Niels Thykier <niels@thykier.net>

To: Anthony Towns <aj@erisian.com.au>, 709460-done@bugs.debian.org

Subject: Re: Bug#709460: [Britney] binNMUs plus source take-over issue

Date: Sun, 26 Apr 2015 18:24:49 +0200

On 2015-04-26 01:12, Anthony Towns wrote:
> Hi,
> 
> Attached is a possible patch for bug#709460. Works against both normal and
> live-data test cases by the looks.
> 
> [...]
> 
> The most notable change is that I pulled the _compute_groups function out
> as the first thing that doop_source does, and then reuse its output later
> rather than working out which binary packages are valid independently.
> (I've left in some commented print statements that add some output as to
> what binaries doop_source is actually looking at which seemed helpful)
> 
> Cheers,
> aj
> 

Thanks for the patch, it has been merged and is now deployed live in
time for the first run in Stretch! :)

Thanks,
~Niels

Send a report that this bug log contains spam.