Re: [boost] [filesystem] path.cpp fails to compile underclang/Cygwin

Subject: Re: [boost] [filesystem] path.cpp fails to compile underclang/Cygwin
From: Beman Dawes (bdawes_at_[hidden])
Date: 2015-01-16 18:27:32

On Fri, Jan 16, 2015 at 4:56 PM, Peter Dimov <lists_at_[hidden]> wrote:
> Edward Diener wrote:
>
>> I have never tried using clang with Cygwin...
>
>
> To clarify, I'm not using clang _with_ Cygwin, that is, I am not using my
> own compiled clang with an existing Cygwin installation. I am using Cygwin's
> clang package, as installed by the Cygwin installer.
>
> Cygwin's g++ was for a long time stuck at 3.4.6, which made it not
> particularly viable for Boost testing, but it's on 4.8.3 now

Better yet, 4.9.2 has been available for a few days now!

>and I find it
> very useful, because it's a POSIX platform, which is good for test coverage.

+1.

For stuff that I am worried about, I run Linux in a virtual machine,
but for every day tests cygwin is great.

> And it has -std=c++11. And clang. What's not to like? Only the fact that
> gcc.archive fails for some reason, which breaks all tests that want to
> statically link to Boost.Test.

I'm having failures for gcc on Windows with the three or four
Filesystem test that link statically. It looks to me like Boost.Build
is missing dependencies, so compiles aren't run soon enough and then
the archive fails. So I'm wondering if the same problem is affecting
Boost.Test. Boost.Timer is also failing gcc static links, but I
haven't had a chance to look at that in any detail.

> And that Boost.Test fails with -std=c++11 due
> to absence of mkstemp and putenv. But apart from that, it's perfect. :-)

I've run into the putenv problem, too. I fixed if for -std=c++11 but
now I'm getting a failure for c++ 03 builds, so I've got to reinstate
a fix for that case.

--Beman