Subject: Re: Building a snapshot and awk
To: Jed Davis <>
From: James Chacon <>
List: tech-toolchain
Date: 12/02/2001 03:34:40
My guess is that by not recompling libgcc one would still beleive they're
broken even if you compile code with -msoft-quad-float. On the surface 
a double -> int won't do quad math, but gcc is internally promoting to that
level inside the libgcc routines which is where the real problem lies.


>James Chacon <> writes:
>> The problem is awk does a lot of double -> int conversion internally and
>> those pass through libgcc routines to do the actual conversions. Some of
>> these promote to quad math first at which point we'll lose.
>For what it's worth, PR 14101 is partially wrong on this: it claims
>that even with -msoft-quad-float enabled, double -> size_t casts are
>broken, but I can't reproduce the brokenness (that is, it works
>correctly for me) with the sparc64 2001082# snapshot.
>#!/usr/bin/perl -- ## "But life wasn't yes-no, on-off.  Life was shades of gray,
>sub f{(($n,$d,@_)=@_)?(substr(## and rainbows not in the order of the spectrum."
>" ExhortJavelinBus",$n&&$d/$n,1),$n?f($d##   -- L. E. Modesitt, Jr., _Adiamante_
>%$n,$n,@_):&f):("\n")}print f 1461,10324,55001,444162,1208,1341,5660480,79715997