PATCH RFA: Only define USE_COLLECT2 in host makefile fragments

[Zack Weinberg]

Ian Lance Taylor <ian@wasabisystems.com> writes:
...
> In this patch, I do the following:
>
> 1) Rename the Makefile variable USE_COLLECT2 to COLLECT2.  This
>    eliminates the confusion between the Makefile variable and the C
>    preprocessor macro, also named USE_COLLECT2, with a different
>    meaning.
>
> 2) Initialize COLLECT2 to @collect2@, set by the configure script.
>
> 3) Add code to config.host to set host_collect2.
>
> 4) Make configure.ac set collect2 based on host_collect2, and
>    substitute collect2 into Makefile.
>
> 5) Eliminate instances of USE_COLLECT2 in Makefile fragments.
>
> 6) As before, give an error if COLLECT2 is empty (i.e., host can not
>    build collect2), and host_xm_defines includes USE_COLLECT2 (i.e.,
>    target requires collect2 to handle global constructors and
>    destructors).

I like this approach.  I would like to hear from one of the
build-system maintainers though.

...
> +#
> +#  host_collect2	Set to yes normally, or no if the host can not
> +#			build collect2

The phrasing here makes it sound like one should set host_collect2=yes
in most stanzas.  Suggest just "Set this to no if the host cannot
build collect2."

zw