Re: current glibc vs debian sid

[Franz Sirl]

At 14:40 22.10.2002, Akim Demaille wrote:
> >>>>> "Roland" == Roland McGrath <address@hidden> writes:
>
> Roland> The coding standards say, "The configure script should also
> Roland> take an argument which specifies the type of system to build
> Roland> the program for."
>
> It would be great if the standards could be updated in this regard.
> There is way too much confusion coming from the existence of *two*
> means to specify the various architecture.  The one based on options
> being clearly superior (more intuitive and more control), the other
> one should fade away.

Well, besides the coding standards problem, the main problem is that you 
can't emulate the old syntax with the new syntax and that the emulation of 
the old syntax isn't working correctly. So either the old syntax 
compatibility has to be fixed or it has to be rejected completely and the 
coding standards need updating then.

Fixing the new syntax requires a documented way to set $host_alias and 
$target_alias for a native build without triggering a cross-compile 
(doesn't auto-set $program_prefix, $ac_tool_prefix, etc.).

What always puzzles me on that issue is what is so difficult about deciding 
whether we have a cross or native build? After setting $canon_build, 
$canon_host and $canon_target (I'm using the canon_ prefix here cause the 
naked $build, $host, $target are documented as deprecated) the decision is 
simply:

        if "$canon_build$canon_host" == "$canon_target$canon_target"
          cross_compile=no
       else
         if "$canon_host" == "$canon_target"
            cross_compile=simple
          else
            cross_compile=canadian
          fi
        fi

After that $cross_compile should be displayed to the user and all 
auto-settings/uses of variables like $program_prefix and $ac_tool_prefix 
should be keyed off it.

Franz.