Open codenaugh opened 3 years ago
I have the same problem on Tribblix (illumos). I think the problem is here:
In my case, rather than something like "solaris" or "sunos", target-os is "unix", which doesn't seem right. I imagine that would fall through to the default case, and then try to link with libutil because I'm using gcc.
pkgsrc has also encountered this problem, it looks like. They just hack around it.
I ended up hacking around it as well by adding case uni* : return <toolset>gcc:<library>rt ;
above the sun/solaris target_os
check on line 630 of tools/build/src/tools/python.jam
I think this would be best reported to the boost.build project, which maintains this logic. I'm unfortunately not familiar enough to offer advise, or even provide patches. Thanks !
Well that's where it was originally reported, and they moved it to boost/python. The issue stems from the target_os
variable getting set to unix
instead of solaris
or sunos
. This causes it to match the last case instead of the first one in tools/build/src/tools/python.jam
, which adds libutil
to the link tree and libutil ddoesn't exist on solaris. I was unable to figure out where and how the target_os
variable is being set incorrectly, as that's the real fix.
That being said, I do not mind if it gets moved back to boost/build
My understanding is that Solaris has no such library, but boost does not have a check to determine whether or not to link it. It just always does.