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

Re: mono_1.1.18-3 (arm/unstable): FTBFS: SIGSEGV



On 2006-11-10 16:21 -0800, Steve Langasek wrote:
> Hi Phil,
> 
> > At a first glance this looks more like a TLS emulation kind of problem
> > than an instruction set discrepancy.  But it'd take a bit more detective
> > work to figure out what's really going wrong, and I guess there is no
> > guarantee that this is the same problem the netwinders are seeing in any
> > case.
> 
> So where does that leave us for etch?  Does a mono that doesn't run on
> netwinders and apparently not on smackdown either, but does run on other
> cats systems, make the grade for release?  If no one has time to investigate
> the problem on netwinder, that at least suggests to me that netwinder isn't
> that important a use case for the porters and probably not the users either.
> 
> Does the "TLS emulation" diagnosis imply a bug elsewhere in the kernel or
> glibc rather than in mono?
> 
> Anyway, our three real choices here are:
> 
> - mono support on netwinder is not RC in the porters' estimation, so the bug
>   can be downgraded or etch-ignored
> - mono support on netwinder is RC in the porters' estimation, but no one has
>   time to work on this problem, so the arm binaries should be removed from
>   the archive for the release
> - mono support on netwinder is RC in the porters' estimation, and there is a
>   porter with the know-how and time to fix this bug who is volunteering to
>   have me nag them once every other day until it's fixed ;)
> 
> If we are still missing information for the porters to decide whether this
> should be RC, what can I do to help get that information?

OK. I don't have a netwinder to test on (tbm took mine), and probably
don't have sufficient expertise either. This failure is extremely
odd. I think we should etch-ignore this bug, it's not ideal, but
no-one is rushing forward to fix it, and netwinders will become less
important during this release. Their use as buildds is the main reason
it's a problem in practice - can we arrange to only have this built on
the non-netwinder machines? 

Removing mono from arm for this release entirely seems worse than
having a version which does work on some(?)/most(?) hardware.


Wookey
-- 
Aleph One Ltd, Bottisham, CAMBRIDGE, CB5 9BA, UK  Tel +44 (0) 1223 811679
work: http://www.aleph1.co.uk/                 play: http://wookware.org/



Reply to: