|
Boost : |
Subject: Re: [boost] [CrystaX.NET] [log] Test failures
From: Andrey Semashev (andrey.semashev_at_[hidden])
Date: 2015-03-22 08:19:06
On Sunday 22 March 2015 13:57:49 Dmitry Moskalchuk wrote:
> > 1. Compilation and linking succeeds but running the test fails due to
> > missing ___tls_get_addr symbol.
> >
> > http://www.boost.org/development/tests/develop/developer/CrystaX-NET-apile
> > vel-19-x86-log-clang-linux-3-5-attr_attribute_set-variants_.html
> This is known problem, caused by absence of proper TLS support in clang
> (x86-only), included to the CrystaX NDK. We're working on this.
>
> > 2. Linking fails due to missing symbols from iostreams for long double.
> >
> > http://www.boost.org/development/tests/develop/developer/CrystaX-NET-apile
> > vel-21-x86_64-log-clang-linux-3-5-attr_attribute_set-variants_.html
> This is known problem too. Again, this is defect of clang, included to
> the CrystaX NDK, which don't yet fully support long double on 64-bit
> targets (there was significant change on switch from 32-bit to 64-bit
> Android in how long double is represented; gcc was adjusted properly,
> but there are still bugs in supporting that in clang). We're going to
> fix this too.
>
> Both issues are not very critical though - just because a) gcc
> toolchains don't have such problems, b) x86 is very rare case for
> Android and c) there is only one Android device on the market, running
> 64-bit Android - Nexus 9. Of course, we're going to fix this anyway, but
> this is definitely not the blocker.
Ok, thanks for the info. Should I report this to CrystaX NDK bug tracker?
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk