|
Boost Users : |
Subject: Re: [Boost-users] M68K-Coldfire: Boost 1.44 seems to have a problem with low level locking
From: Vladimir Prus (vladimir_at_[hidden])
Date: 2010-09-15 05:07:44
Andreas Wehrmann wrote:
>
>> Boost is now using gcc atomics, but it assumes them to be available everywhere, which is not the
>> case. Adding
>>
>> define=BOOST_SP_DISABLE_THREADS
>>
>> or
>>
>> define=BOOST_SP_USE_PTHREADS
>>
>> to build command should fix this error (at the price of either thread safety or performance)
>>
>
> I tried BOOST_SP_USE_PTHREADS and BOOST_SP_USE_SPINLOCK of which neither
> one works.
> I am not willing to pay the price for losing thread safety, which is why
> I haven't tried the other define.
>
>>> Btw. I'm also using boost version 1.38 in another project (but on the
>>> same platform).
>>> My compiler is g++ version 4.2.0 20070318 (prerelease) (Sourcery G++
>>> Lite 4.2-35).
>> That's pretty old compiler, our current lite releases are 4.4-*. However, upgrading is not
>> likely to fix your immediate issue.
>>
>> - Volodya
>
> I know, we upgraded to the 4.4 version once and had nothing but problems
> which is why we reverted to 4.2.
>
> Thanks for your hints thus far,
> do you have any idea of why BOOST_SP_USE_PTHREADS doesn't work?
What is sgu_loopback_timer.o? If this is your file, it also should be compiled
with that define.
If you're using 1.38, it might have a bug whereby define=XXX is not
respected on the command line. Hacking boost/smart_ptr/detail/sp_counted_base.hpp
may be more reliable.
I am not entirely sure whether hacking user.hpp will work or not.
HTH,
Volodya
Boost-users list run by williamkempf at hotmail.com, kalb at libertysoft.com, bjorn.karlsson at readsoft.com, gregod at cs.rpi.edu, wekempf at cox.net