On Sat, Dec 20, 2025 at 10:15:35AM +0100, Fabian Groffen wrote: > On 19-12-2025 22:06:06 -0500, Mike Gilbert wrote: > > On Fri, Dec 19, 2025 at 9:58 PM Mike Gilbert <[email protected]> wrote: > > > > > > On Fri, Dec 19, 2025 at 3:23 PM Fabian Groffen <[email protected]> wrote: > > > > > > > > Use the proper methods on Darwin and Solaris hosts to get available > > > > memory and available swap in kibibytes. > > > > > > You should probably check CBUILD instead of CHOST; if we are > > > cross-compiling, you care about the system the software is being built > > > on, not the system we are targeting. > > Ack, makes sense. > > > Or maybe check uname -s? This eclass really has nothing to do with > > the toolchain, so checking toolchain variables seems weird. > > I thought CHOST and CBUILD are defined by profile/environment, not > toolchain eclass or something, right? The package manager defines > CBUILD when not set, so it's always there > (lib/portage/package/ebuild/config.py).
Believe meant toolchain variables in the sense of being toolchain
related rather than from the toolchain eclasses.
That aside, the package manager is not required to set CBUILD, the
PM spec has a "or be unset" in:
"CHOST, CBUILD and CTARGET, if not set by profiles, must contain either
an appropriate machine tuple (the definition of appropriate is beyond
the scope of this specification) or be unset."
So typically when we read CBUILD, we use ${CBUILD:-${CHOST}}.
... and personally think it's ok to use here.
--
ionen
signature.asc
Description: PGP signature
