GCC build failed with your patch on 2001-01-09T11:35:00Z.

[Richard Henderson]

On Tue, Jan 09, 2001 at 02:13:25PM -0800, Mark Mitchell wrote:
>     Richard> The class of targets for which I consider the patch to be
>     Richard> incorrect are 32-bit natives for which there is a runtime
>     Richard> selectable 64-bit variant -- PA, MIPS, and PowerPC.
> 
> Why is the patch incorrect for these targets?

Well, not "incorrect" but "undesirable" since it slows them down
compile times unnecessarily.  A 32-bit H_W_I is adequate until one
uses one of the target switches that selects the 64-bit variant.

Of course, this highlights that Sparc's solution is really the
best in the short term.  Not to support the 64-bit target at all
until the host integer size is wide enough.

> Isn't bitsizetype still 128 bits wide here?

No, bitsizetype is chosen at compiler runtime.  It is DImode for
these 32-bit targets.


r~