On 07/05/2020 14:48, Paul Goyette wrote:
No I think that just causes the build system to set up a python3 environment instead.Host system is amd64 running NetBSD 9..99.59, and pkgsrc is updated totoday. I've rebuilt all dependencies from scratch into an initially sandbox, and now I'm getting a repeatable build error for firefox.The first error message seems to be psutil/_psutil_common.c:9:10: fatal error: 'Python.h' file not found #include <Python.h> ^~~~~~~~~~ 1 error generated. error: command 'gcc' failed with exit status 1 but it seems that this is not critical:
This matches what I see on 9-stable on lots of different files. The compiler is not picking up on the wrapped standard C++ headers.In file included from /tmp/pkgs/www/firefox/work.x86_64/build/dist/include/mozilla/Array.h:14:/tmp/pkgs/www/firefox/work.x86_64/build/dist/system_wrappers/ostream:3:15: fatalerror: 'ostream' file not found #include_next <ostream> ^~~~~~~~~ There are multiple files mentioned in these errors, including "utility", "type_traits", "cmath", and "new".
It's not specific to this file. Just went and read the gcc documentation for #include_next. Its like #include but only looks in the include path AFTER the directory container the header that contains the #include_next so if the directory that contains the system headers is before the wrapper header file in the path that would trigger this. Its also possible that clang interprets include_next differently.These errors all seem to be related to compiling "wrappers.cpp".
So I think it boils down to the order in which the various include directories get passed to the compiler. :( I can compare command lines to my last working build later and see if I can see what is going on.
Mike