Bug#685209: unblock: ball/1.4.1+20111206-4
Hi everyone,
First of all, thank you for your comments! I'll try to clarify some
things, but I am not sure yet if I understood everything correctly.
First, about the change of the maintainer email. I was absolutely
convinced that I'd put that in the changelog, but as I just noticed, I've
forgotten to do that in the hurry of the last few days.
Then, about the patches. Concerning the bond order stuff (patches 0007,
0008, and 0009), this was a difficult decision. In upstream, we first had
a rewrite of the bond order code and then fixed the gcc problems. It would
be relatively simple to write a small patch based on the old version of
the code that gets it to *compile* with gcc 4.7. However, the new compiler
exposed some memory issues in the code that made the code *crash* when
running a gcc 4.7 - compiled version. So I've tried to isolate, from the
rewrite in upstream, those parts that fixed the problem. The rewrite in
upstream was unfortunately not really granular, so getting this stuff
untangled turned out to be a bit of a nightmare. Sorry, I should have
mentioned that in the changelog, but the last days were a little hectic.
Now, when I got the package to compile on testing with gcc 4.7, I noticed
that a change in sip also broke the build. So I've added a quick patch to
change that (0011), but probably also forgot to mention that in the
changelog. Sorry about that!
Best regards,
Andreas
Am 18.08.12 13:52 schrieb "Steffen Möller" unter <steffen_moeller@gmx.de>:
>Dear Adam,
>
>On 08/18/2012 01:34 PM, Adam D. Barratt wrote:
>> On Sat, 2012-08-18 at 13:17 +0200, Steffen Möller wrote:
>>> On 08/18/2012 12:48 PM, Adam D. Barratt wrote:
>>>> On Sat, 2012-08-18 at 12:05 +0200, Steffen Moeller wrote:
>>>>> The uploader is upstream who kindly added a patch to fix a FTBFS.
>> [...]
>>>> While I'm happy to believe that a number of the patches are required
>>>>to
>>>> fix building with gcc-4.7, these at least don't look like they are:
>>>>
>>>> patches/0007-Fixed-the-FPT-version-of-bond-order-assignment.patch
>>>> | 447 ++++
>>>> patches/0008-Added-MAX_PENALTY-option-to-bond-order-assignment.patch
>>>> | 975 ++++++++++
>>>> patches/0009-Fixed-a-bug-in-the-AssignBondOrderProcessor.patch
>>>> | 271 ++
>>>>
>>>> Indeed some of those look like they may well introduce ABI changes if
>>>> the changes are reflected in libball.
>>
>> Some comment on this, possibly from upstream, would be helpful. I
>> realise that nothing outside of ball uses the libraries, but they (I
>> missed libballview) are shipped as public soversioned libraries and
>> nothing in the package dependencies appears to ensure that the packages
>> are upgraded in step.
>>
>>>> This obviously isn't a gcc 4.7 fix:
>>>>
>>>>
>>>>patches/0011-Fix-compilation-of-Python-bindings-with-new-sip-vers.patch
>>>> | 25
>>>>
>>>> It may well be needed for the package to build in unstable now, but
>>>>it's
>>>> certainly not covered by "Fix compilation with g++ >= 4.7".
>>>
>>> >From the description of those bugs I tend to agree that indeed that
>>>is more
>>> than the FTBFS required. Upstream mostly works on version 2 of BALL
>>>now, so
>>> knowing whatever ships with Wheezy now to be worked with for a while,
>>>I presume
>>> Andreas (upstream) to just have wanted some of the later experiences
>>> backported. A respective description in the changelog is missing.
>>
>> At the very least, yes.
>>
>>>> Was simply forcing building with gcc 4.6 considered as an option?
>>>
>>> I did not know myself this was allowed for anything in stable. The
>>>bond-order
>>> changes were apparently important. With my Debian Med hat on, I am
>>>always
>>> eager to have such close ties with upstream so Debian gets the best
>>>that is
>>> possible, not only something that compiles.
>>
>> In general, I'd agree. The focus on changes during a freeze is somewhat
>> narrower, however.
>>
>>> I understand that you see difficulties to accept the package for Wheezy
>>> as it is. Would it help you to have the debian/changelog updated with
>>>the
>>> maintainer change and a description about those changes to the bond
>>>order?
>>
>> A description of those patches would certainly help. As would an
>> explanation as to why they're so important that they either meet the
>> published freeze guidelines or merit an exception.
>
>If you were the ftpmasters then I would now have suggested to reject the
>upload for Andreas to upload a corrected version. Actually I had already
>formulated that sentence when I realised that the package already is in
>unstable :o/
>
>@Andreas: Since everything is already uploaded and went through the
>buildds and was shipped to all the mirrors and your pending explanations
>now are mostly for the release team to decide, please first formulate
>an email to Adam outlining your reasoning on those non gcc
>4.7-associated changes. Trusting you in your initial reasoning
>I have some remaining hope for the package to be acceptable nonetheless,
>at least we/you will learn if the bond-bits can remain in or if you rather
>have the version currently in testing removed because of those bond issues
>you now know about.
>
>Many thanks to Adam for his constructive review and regards to all
>
>Steffen
Reply to: