Subject: Re: [Boost-build] Building Boost with MinGW
From: Steven Watanabe (watanabesj_at_[hidden])
Date: 2010-07-12 12:54:24
Peter Rockett wrote:
> On 12/07/10 16:08, Steven Watanabe wrote:
>> Peter Rockett wrote:
>>> Newbie question: I am trying build Boost with MinGW using bjam from
>>> inside MSys, the Bash-like interface supplied with MinGW.
>>> When I invoke bjam with the "--toolset=mingw" option (as detailed in
>>> Table A.1 of the Boost.Build docs),
Table A.1 is for building bjam itself. It doesn't apply to using
>>> I get a set of error messages like "mingw.jam: No such file of
>>> directory" and "rule mingw.init unknown in module toolset", all of
>>> which imply that "mingw" is not a valid option.
>>> How do I build Boost with MinGW?
>> Use toolset=gcc.
> Steven - is that a definitive instruction or give-this-a-try advice?
gcc is the correct toolset for mingw.
> I have tried gcc but I am not sure it works. The math lib names end-up
> decorated with "c99" - MinGW does not conform to c99. MinGW is not
> actually gcc - it doesn't use glibc.
The math libraries have c99 because they are implementations of the
c99 math functions, not because they use c99.
> Actually, I am trying to build a library (libDAI) which uses Boost.
> libDAI builds 'out-of-the-box' on Linux but with MinGW, the build
> fails with an obscure problem. My first suspicion is an issue with my
> MinGW-built Boost libs. (Could be an issue with libDAI/MinGW, I
> accept. But I'd like to be clear that's my problem, not something with
> the Boost libs.)
> I have also tried toolset=gcc-no-cygwin (which passes the -no-cygwin
> flag to gcc thereby invoking the mingw interface). This seems to
> produce warnings about c89 not supporting long doubles (which I'd
> expect with MinGW) but the libDAI build still breaks.
> Regardless, what has happened to the toolset=mingw option?
I don't think it ever existed.
Boost-Build list run by bdawes at acm.org, david.abrahams at rcn.com, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk