source-highlight: fails with dyld: lazy symbol binding failed

[skymoo (Adam Mercer)]

source-highlight, dependency of gtk-doc since r124901, fails to build with:

:info:build dyld: lazy symbol binding failed: Symbol not found: __ZN5boost9re_detail12perl_matcherIPKcNSt3__19allocatorINS_9sub_matchIS3_EEEENS_12regex_traitsIcNS_16cpp_regex_traitsIcEEEEE14construct_initERKNS_11basic_regexIcSC_EENS_15regex_constants12_match_flagsE

Full build log attached.

Seems to be related to boost, I have the following version installed:

 boost @1.56.0_1+gcc48+no_single+no_static+python27 (active)

[neverpanic (Clemens Lang)]

You're on a system that uses -stdlib=libc++ – I don't think you can have a boost built with GCC. We should remove the GCC variants from boost.

[ryandesign (Ryan Carsten Schmidt)]

Yes, except that e.g. on PowerPC, boost blacklists all Xcode compilers, and clang won't build on PowerPC either, so the gcc variants are the only way to get boost to build on PowerPC.

[skymoo (Adam Mercer)]

Replying to cal@…:

You're on a system that uses -stdlib=libc++ – I don't think you can have a boost built with GCC. We should remove the GCC variants from boost.

If that's the case why does the boost port let me enable a gcc variant on such a platform?

[neverpanic (Clemens Lang)]

Replying to ram@…:

If that's the case why does the boost port let me enable a gcc variant on such a platform?

As I was saying: We should remove them. We can keep them on PowerPC platforms, but they shouldn't be available on any system that uses libc++.

[neverpanic (Clemens Lang)]

Has duplicate (different symptom, same problem) #44856.

[ryandesign (Ryan Carsten Schmidt)]

Removed the compiler variants from boost in r125939.

[mf2k (Frank Schima)]

Is this still a problem for you?

[skymoo (Adam Mercer)]

No it looks like removing the GCC variants from boost did the trick.