From: Johan Knol <johan.knol@id...> - 2001-10-09 14:49:59
This probably isn't an endian issue.
I revived my Sparc5/Solaris2.6 box because it bothered me.
Some statistics: ftp sdcc.tar; configure; make
Solaris 2.6/Sparc5: 732 kb/s; 4m12s; 38m12s
Linux RedHat 6.0/P2-300: 595 kb/s; 31s; 5m28s
} so I'd rather not do that again;
The problem is the "long long" type. We need that to distinguish e.g.
unsigned/signed long 0x80000000. A problem I found is that on Solaris the
size specifier 'L' in a scanf format string is ignored for non-float types,
whereas linux (mingw? ppc?) treats it as an equivelant to 'll'. I think msvc
(do we still support that) even expects 'I64', don't know about Borland.
Anyway, it's 'll' for now. I hope that will help the PPC build.
----- Original Message -----
From: Johan Knol <johan.knol@...>
Sent: Tuesday, October 09, 2001 3:24 PM
Subject: Re: [sdcc-devel] Automated build output (ppc-unknown-linux2.2)
> > gen/z80/muldiv/muldiv_storage_none_attr_none_type_char.out:13:--- FAIL:
> "Assertion failed" on i/-4 == -25 at
> This must be an endian issue. I know there is a possibility to compile and
> debug on a sourceforge ppc machine. Someone has some short hints how to do
> that? Thanks.
> Hmmm, int=32. All the distinctions made in valWhateverOperation() between
> long an int are useless because long==int. Hmmmmmmmm
Get latest updates about Open Source Projects, Conferences and News.