Host Versus MacPorts lib[std]c++ [was: Re: Fortran recipe]

[Jeremy Huddleston Sequoia]

On Aug 29, 2013, at 7:12, Michael Dickens <michaelld at macports.org> wrote:
> On Aug 28, 2013, at 6:06 PM, Jeremy Huddleston Sequoia
> <jeremyhu at macports.org> wrote:
>> One option is to have the gccXX ports use 'clang -c -x assembler' instead of as.
> 
> A quick search for clang being explicitly blacklisted shows around 175
> ports out of roughly 17500 total ports; roughly 140 explicitly blacklist
> some form of gcc. Not all of these ports use C++, and some are probably
> unmaintained/obsolete and can thus be dropped, which drops the total %
> somewhat below 1% for either category. Hence, it is fair to believe that
> the vast majority (some 99%) of all MacPorts ports can use clang or gcc
> interchangeably, and hence can the vast majority can also use libc++ for
> the C++ runtime library.

Yeah, I've been auditing the blacklisting and explicit configure.compiler setting on and off for the past couple years.  With changes in base 2.2 and Ryan's compiler_blacklist_versions PortGroup, it is significantly easier to express exactly what does and does not work, so I've been trying to clean those up and note in the Portfile what the underlying problem is.  Without actually reading through, I believe the exact number of C++ ports that have clang blacklisted is probably in the 10s to 20s.

> Given the prior discussion and the above: It seems to me like creating a
> branch (like what Mojca just did for the WX stuff) that sets
> "configure.cxxflags-append -stdlib=libc++" when compiling with clang,
> and "AS='clang -c -x assembler'" (or, whatever works) when compiling
> with GCC would be an interesting experiment.  Those of us interested in
> having a common C++ runtime could contribute to it, try to get as many
> ports as possible tested and working with this methodology.  If we're
> not going to shoot for a common libstdc++ (for any reason, not just
> because certain developers cannot touch that code), then I'm all for
> shooting for a common libc++ -- and, sooner is better than later IMHO.

Ok, I'll do something like that in a few months once I'm satisfied that we're ready for it.

--Jeremy

-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 4145 bytes
Desc: not available
URL: <http://lists.macosforge.org/pipermail/macports-dev/attachments/20130829/e0e14868/attachment-0001.p7s>