RE: State of TOOLSRC?

At 11:13 Uhr +0100 5.3.2003, Dittmar, Daniel wrote:
>The first compiler where there were actual problems was gcc 3.1 for IA64
>and they mended their ways. Compilers for Solaris, AIX, HP-UX and DecUnix
>all use sizeof (bool) == 1, otherwise, we would have fixed that long ago.
>It looks as if we have to anyway.

NetBSD 1.6 ships with gcc 2.95.3, and MacOS X 10.2 ships with a hacked gcc
3.1 (apparently ppc support is nominal in gcc 2.xx). So it looks like gcc
in general uses whatever the CPUs natural word size is for bool.

-- Of course, when I came across the configure error, the reason wasn't
obvious, and I googled around a bit. There were a few reports about
problems with Alpha, too. Like with 'int' in the  case of C, a C++ compiler
is free to choose the target machine's natural word size for bool. i386
CPUs do fine with char because of their i8080 grandsire, but for RISC CPUs,
byte accesses are inefficient.

>Although, if I remember right, the places
>that need fixing are far from obvious.

I can imagine.  ;)

>> So, TOOLSRC looks like it could use some work. Has the cvs source state
>>ever built? Has it built on any cpu besides i386? On any OS besides a
>>linux
>
>It's been build some time ago on Solaris/Sparc and AIX and others have
>been able to build them on Solaris/I386 and Linux/s390.

With the vendor compilers or with gcc?

>But it certainly could use some work, especially as there are still
>platform tests all over the place.

Yep. When I saw

#if u3b || M32 || u3b15 || u3b5 || u3b2 || pyr || sun || ultrix
#define sigvec sigvector
#endif /* several */

I felt young again. Back then, I was still banging on my Sinclair ZX81.   ;)

>We could also use a lot of help on enabling autoconf for the rest of the
>software.

[...]

>sys/src/base contains everything that is created by Makefiles and is
>needed to bootstrap the development environment. A few Perl-Scripts are
>needed as well and they a copied in basescripts.mk.
>
>The rest of the tools is build by calling the SAP DB make tool vmake:
>in Makefile.in

Can you draw the line between what would have to be autoconfiscated, and
where vmake takes over?

	hauke

--
/~\  The ASCII Ribbon Campaign
\ /    No HTML/RTF in email
 X     No Word docs in email
/ \  Respect for open standards