Question about meson and boost libraries

[Benjamin Redelings]

Thanks for all the really helpful comments!  I'm also sticking 
mlog.warning( ) comments into the meson code to find things, and I 
figured out why the link isn't happening.

I have a fix to finding boost libraries with a -mt suffix in meson here:

     https://github.com/mesonbuild/meson/pull/2824

However, even after that fix I have to define BOOST_ROOT=/opt/local to 
find boost.  For manual builds, if don't do PATH=/opt/local:$PATH then 
meson uses pkgconfig from homebrew and runs the wrong pkg-config.  But 
with those options, at last the configure step succeeds.

-BenRI


On 12/22/2017 11:51 AM, Mojca Miklavec wrote:
> On 22 December 2017 at 19:14, Benjamin Redelings wrote:
>> Hi,
>>
>>      I'm new to macports.  I'm the author or the bali-phy software, and I'm
>> working on updating the ancient Portfile for it.  The software has changed a
>> lot since someone wrote the original Portfile years ago, and I was
>> completely unfamiliar with the Portfile format yesterday, so I have a bit of
>> learning to do.
> Welcome to MacPorts :)
>
>>      One issue I'm running into is that meson isn't finding boost at the
>> configure stage.  It might be because its not looking for libraries with the
>> -mt suffix.  However, while I'm continuing to look into this, I thought I'd
>> ask what the general state of using meson with macports is.  Is there a
>> place I can go to see discussion of this?
> Probably the only discussion we ever had is the one that Ken pointed you to.
> The general state is that:
> - We just integrated the PortGroup, support is super rudimentary
> - Meson's treatment of dylibs is pretty much broken (latest version of
> meson recently changed something in that respect, but I don't fully
> understand how that rpath-stuff works, so something needs to be fixed
> in either case).
> - The best place of discussion is this list or irc channel or
> (probably more helpful) the meson mailing list or IRC channel.
>
> If meson cannot find boost, someone (either you or another macporter)
> should try to start fighting with meson or you should ask the
> developers for help. It would help to run it in verbose mode to get
> some hints why meson is failing to find the library (I'm afraid that
> the code is pure python where you cannot really turn debug on with a
> command line switch).
>
> Last time I tried debugging a few things I usually had to inject a lot
> of debug statements into python code. A totally stupid example is that
> if you don't have git installed, it won't find any libraries at all (I
> was trying to use gtest) because it tries to consult git first and the
> failure to run "git info" is interpreted as "module cannot be found".
> I submitted a number of bug reports, but the resolution rate & speed
> is usually quite low if you don't actively participate in debugging
> and fixing (and even if they fix something, the bug reports often stay
> open).
>
> Then again, MacPorts doesn't excell in ticket resolution speed either :(
>
>>      I'm pretty happy with meson in general - I'm finally willing to switch
>> away from autotools.  I've been able to generate cross-builds from linux to
>> mingw-w64 and darwin, and I've done native builds on linux and osx.
> With the exception of buggy wine support I was super happy with
> cross-compiling with MinGW from MacPorts as well.
>
>> But integration with packaging tools seems to be another layer...
> It would help to get some debug info first. If boost needs to be found
> via pkg-config, it should be sufficient if meson finds the correct
> pkg-config. If there's a different magic involved, then I'm not sure
> about the proper way of pointing meson to the prefix of MacPorts.
>
> Mojca