assert in rts/Schedule.c

Karel Gardas karel.gardas at centrum.cz
Mon Jan 9 13:34:06 CET 2012


On 01/ 9/12 11:40 AM, Simon Marlow wrote:
>> Thanks for the information, however mine is previous one:
>>
>> 527
>> 528 ASSERT_FULL_CAPABILITY_INVARIANTS(cap,task);
>> 529 ASSERT(t->cap == cap);
>> 530
>>
>> and also this is on single-core ARM. Interesting...
>
> This could be a problem. I haven't seen any of these in my builds. Does
> it happen on the 7.4 branch or only HEAD?

This is on GHC HEAD. However this was not so "standard" testing here, as 
this was a build which used Ubuntu's provided preliminary LLVM 3.0 
packages. I did this in order to test them. So now, I'm back to my own 
LLVM 3.0 build and building the same GHC tree with this to see if this 
is the problem caused by interaction with either Ubuntu's LLVM 3.0 
packages, or new Ubuntu 11.10 or this is really a bug in HEAD. My bet is 
on Ubuntu's LLVM 3.0 packages as I guess they built them with GCC and 
default optimization which is not that good[1].

Anyway, new build is just finishing and full testing takes about 24 
hours on this board so I'll send you result during Wednesday probably...

Thanks,
Karel
[1]: 
https://ghcarm.wordpress.com/2011/12/30/testing-llvm-3-0-on-ubuntu-11-10-arm/



More information about the Cvs-ghc mailing list