illumos gate

If the AST headers were split off, this becomes a different story... system/header will have no technical reason to depend on header-math, but logically the libm API is just part of the system interfaces (when libm integrates there may not be a reason to have a distinct header-math).

Yeah, when libm integrates I want the math headers in system/header properly.

I'll be ready with LIBM on next week with checks on SPARC.
need check full builds with lint.
another idea - maybe we will provide separate -dev package with development staff ?
will be possible: static libs, headers, .so symlinks, etc ...

Igor S.Kozhukhov wrote:

I'll be ready with LIBM on next week with checks on SPARC.
need check full builds with lint.
another idea - maybe we will provide separate -dev package with development staff ?
will be possible: static libs, headers, .so symlinks, etc ...

One very important feature of pkg(5) uses tagging to indicate development files, these tagged files are optional "facets" of a package. pkg:/system/library could use this to make pkg:/system/header obsolete.

For extra clarity, while system/library could, it won't. The right way to do that is to move content out of system/library and system/header into new, smaller packages which then do the right thing.

The right thing to do with the math headers is either add them to system/header, or to add them to the math library package itself. I would prefer they be in the math library package, now I think about it.

Oh, and yes, what Albert said. No -dev packages.