Re: Bits from the Release Team: Architecture health check

[Niels Thykier <niels@thykier.net>]

On 2014-01-30 03:49, Yaroslav Halchenko wrote:
> 
> On Wed, 29 Jan 2014, Niels Thykier wrote:
>>  * sparc
>>    - We have seen no improvements.  Therefore, out of date
>>      binaries on sparc will no longer prevent packages from 
>>      migrating to testing and Britney will be allowed to break
>>      existing packages in testing on sparc.
>>    - We will review sparc again in about 2 months; should porters
>>      have stepped up and resolved the concerns we had about sparc
>>      we will consider undoing this change.  Otherwise, we will
>>      remove sparc from testing as well.
> 
> FWIW -- from my pragmatic PoV -- quite a few times the problems with
> FTBFS unittests on sparc were legit bugs which were not exercised by
> unittests on more popular architectures (who would test all possible
> boundary conditions, right? ;-)).

Hi,

Indeed, that is great when architectures catch legitimate bugs and there
are porters to help fix them (or, at least, cluebat maintainers into
fixing them).  But when there are no active porters, the port becomes a
chain that drags us down.
  For the record, we (sort of) have a similar problem with poorly
maintained packages - but there we can at least setup an automated
system to remove most of the problems for us.

> That would be pity if sparc goes away :-/
> 

If you feel that way, I recommend that you (bribe someone to) help the
port.  At the current time, there are two major issues with sparc.

 * Make gcc-defaults use at least gcc-4.8 as default compiler.
   - Of course, this presumes that gcc-4.8 works on sparc and does not
     horribly break all builds on the architecture.
 * Have the sparc kernels in stable fixed
   - Currently, DSA are forced to use kernels from oldstable.  Probably
     anyone using sparc/Wheezy are forced to that, but so far we only
     heard complaints from DSA[1].
   - Even if you don't care for sparc in testing, this problem affects
     our ability to support sparc in stable.
   - To be honest, I am impressed DSA haven't just yelled "We veto
     sparc for Jessie until the stable kernels are fixed".

But until that happens, sparc in testing will slowly degrade - so I
recommend you decide and act quickly.


Let it be no secret: If there is no visible progress on salvaging sparc
in the next two months (like there wasn't for ia64 or sparc in the past
two), then I will most likely be on the side encouraging the removal of
sparc from testing.

~Niels

[1] I forgot the actual details of why it is broken.  I believe it was
something down the lines of "doesn't boot at all".  Anyway, I am sure
the DSA can remind me on that if needed be.