Cross compilers without assemblers (rant)

[Daniel Jacobowitz]

On Wed, Oct 20, 2004 at 03:45:19PM -0700, Zack Weinberg wrote:
> 
> $ .../configure --target=powerpc-apple-darwin --enable-languages=c
> ...
> $ make all-gcc TARGET-gcc=cc1
> ...
> checking assembler for .machine directive support... no
> *** This target requires an assembler supporting ".machine"
> make: *** [configure-gcc] Error 1
> 
> Yep, I don't have an assembler available that supports .machine.
> Matter of fact, I don't have an assembler available that supports
> Mach-O.  Or the PowerPC instruction set.  And I DON'T CARE, because
> all I'm trying to do is make sure I can still compile cc1.
> 
> I deeply resent having to jump through extra hoops for a testing
> process which has already chewed up most of the afternoon.
> 
> </rant>
> 
> zw
> 
> (what, a constructive suggestion you want?  Fine.  Don't error out in
> this circumstance when host != target.  Better still, don't even try
> to probe features of the assembler when host != target, <target>-as
> doesn't exist, and this isn't a unified-tree build; just assume the
> user knows what they're doing.)

Then please require --enable-I-know-what-I-am-doing.  I just wasted
most of a day tracking down a problem with its roots in the assembler
feature testing, and having these things fail at GCC build time instead
of forty packages later would be a huge relief to me.

-- 
Daniel Jacobowitz