GCC build failed with your patch on 2001-01-09T11:35:00Z.
David Edelsohn
dje@watson.ibm.com
Tue Jan 9 11:12:00 GMT 2001
>>>>> Alexandre Oliva writes:
Alexandre> Then you must arrange for MAX_LONG_TYPE_SIZE to be #undefined or
Alexandre> #defined to 32 (as long as LONG_TYPE_SIZE isn't wider than that).
Alexandre> AFAICT, MAX_LONG_TYPE_SIZE is defined to 64, and that's the reason why
Alexandre> you get a 64-bit HOST_WIDE_INT.
The rs6000 port supports both 32-bit and 64-bit computation based
on a runtime flag. Therefore, the MAX_LONG_TYPE_SIZE is 64 because it can
attain that value with the appropriate runtime flag. This is necessary
for 64-bit support.
I was unaware of your patch until this discussion arose. I think
that it is a very bad patch and I urge you to revert it. hwint.h should
not be assuming what size HOST_WIDE_INT to use other than the maximum,
natural host wide int. If you want to allow someone to use "long long",
then make that a configure option, not imposing that on every target by
default.
The rs6000 port cannot set MAX_LONG_TYPE_SIZE to 32. If you
insist on retaining the patch, then you need to fix 64-bit host support
for the rs6000 port.
David
More information about the Gcc-regression
mailing list