Re: [Boost-build] Meaning of gcc-mingw toolset

From: pbristow_at_[hidden]
Date: 2020-04-16 14:11:13

> -----Original Message-----
> From: Steven Watanabe <watanabesj_at_[hidden]>
> Sent: 16 April 2020 12:04
> To: Edward Diener via Boost-build <boost-build_at_[hidden]>
> Subject: Re: [Boost-build] Meaning of gcc-mingw toolset
> 
> AMDG
> 
> On 4/15/20 4:45 AM, Edward Diener via Boost-build wrote:
> > In the multiprecision testing jamfile there is a line which has the
> > requirement:
> >
> > <toolset>gcc-mingw:<cxxflags>-Wa,-mbig-obj
> >
> > I do not see a toolset called gcc-mingw nor do I see a subfeature of
> > the gcc toolset called mingw.
> 
> The subfeature is called flavor.
> 
> > I do realize that the gcc toolset has support for mingw gcc on
> > Windows, which I use myself in my user-config.jam with the plain
> > designation of:
> >
> > using gcc : nnn : etc ;
> >
> > but the nnn is always a version number, like 9.3. The gcc.jam picks up
> > that this is mingw-w64/gcc implementation internally and works fine
> > with it. But whatever '<toolset>gcc-mingw' is supposed to mean it does
> > not applu to my "toolset=gcc-nnn' when I run the test.
> 
> What is the path component for the toolset?  It should be gcc-mingw-nnn.
> Otherwise the detection for mingw isn't working.
> Also, I recall that there used to be problems if you have both mingw and non-
> mingw versions of gcc.  I don't remember whether that was ever fixed.

Thanks for this enlightenment.  I think that not understanding how this should work was part of our problem(s).

Also get fails for a multiprecision example floating-point_examples.cpp that used to work before.

But now I see some reports that suggest that there might also be a problem with mingw and option -mbig-obj' -
not only has GCC version changed but also a new release of mingw64.

And none of the test runners using mingw, so we have no warning of trouble.

But armed with your invaluable info, we may be able to pinpoint the faults more intelligently.

Thanks

Paul