[cfarm-users] Some issues with cfarm215

[Jonathan Wakely]

On Fri, 23 Jan 2026 at 14:23, Peter Gutmann via cfarm-users
<cfarm-users at lists.tetaneutral.net> wrote:
>
> Denis Ovsienko via cfarm-users <cfarm-users at lists.tetaneutral.net> writes:
>
> >stress-ng uses an original make-only build system, I did not try it on
> >Solaris, but on Haiku it worked good enough to fix the few broken bits. It
> >detects OS features like Autoconf, but faster, before the main build, and
> >what it cannot detect it expects the user to specify in environment variables.
>
> That's what my code currently does, it tries to autoconfigure itself for every
> target environment, which is why the shell script just for figuring out
> compiler options is nearly 2,000 lines long (admittedly a lot of that is
> comments, where the term "braindamage" features in several).  Particularly
> entertaining is when the compiler reports via --help that it supports foo but
> when you use foo it says it's not supported, but there are many more.  And
> tricky are the ones where compiler options then affect linker options, so if
> you specify CFLAGS=x then you also need LDFLAGS=y to match.
>
> So the question was really, are some of the unusual configs on the cfarm
> systems one-offs, or likely to be found in a lot of other systems out there?

GCC using the Solaris linker is not "unusual" it's the recommended way
to install it on Solaris and has been for as long as I can remember
(the docs have explicitly recommended it since at least 2008).

The cfarm tries to *avoid* unusual setups, what you get is the OS as
it's intended to be used.

> In other words can I build with setting CFLAGS=x / LDFLAGS=y just for that
> system or do I have to modify the configure scripts to handle even more
> options?

This is too vague to know how to answer.