I just tested the new binaries with DMD/travis-CI and of course the build failed immediately:

https://travis-ci.org/dlang/dmd/jobs/182649710

It looks like this is a binutils bug: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=57017

So I rerun the test with the sysrooted toolchain (this toolchain includes binutils but doesn't link any system installed libraries without a -L/usr/lib flag) and this worked fine:

https://travis-ci.org/dlang/dmd/jobs/182653194


So what shall we do about this now? Simply include binutils into the x86_64 native compiler toolchain? I'm not sure if this has any drawbacks compared to using the system linker... I could probably build a 4.9.3 instead of a 4.9.4 compiler as well... None of these solutions sound nice though.

Am Fri, 9 Dec 2016 18:15:56 +0100
schrieb Johannes Pfau <nospam@example.com>:

> I just tested the new binaries with DMD/travis-CI and of course the build failed immediately:
> 
> https://travis-ci.org/dlang/dmd/jobs/182649710
> 
> It looks like this is a binutils bug: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=57017
> 
> So I rerun the test with the sysrooted toolchain (this toolchain includes binutils but doesn't link any system installed libraries without a -L/usr/lib flag) and this worked fine:
> 
> https://travis-ci.org/dlang/dmd/jobs/182653194

I probably sent this message too early, looks like we do have some kind of regression in GDC here as well. At least the DMD build passes the ice8511 test: https://travis-ci.org/dlang/dmd/jobs/182653192