pkgsrc-Users archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

XSB Prolog on NetBSD (Re: wine-devel hangs with high CPU usage)



Mayuresh <mayuresh%acm.org@localhost> writes:

> On Sun, Nov 13, 2011 at 04:45:14PM +0400, Aleksej Saushev wrote:
>> >
>> > I can try it as long is does not require too much of insight into the app
>> > itself.
>> 
>> WINE Is Not Easy.
>
> Oh yes. I am pretty sure it isn't. Hence what I meant was things like
> tweaking of options that must be tweaked on NetBSD etc. (An example below
> will clarify what I mean by those tweaks.)
>
>> You started talking about WINE, but it seems that what you need actually
>> is some open-source Prolog interpreter that builds (and potentially works)
>> for FreeBSD and Linux. What we need to know now is its name. (Is it Ciao? 
>> XSB?)
>
> Sorry about that. I was citing an example where possibly a few tweaks
> could get it work on NetBSD. I was looking for similar guidance on wine.
> Hence did not elaborate on Prolog.
>
> Yes, it's XSB. Following is what I could and couldn't do:
>
> The XSB configure script does not recognize NetBSD. I just added netbsd as
> orred pattern to freebsd (essentially made it do the same things it does
> for freebsd including #define FREEBSD etc. for netbsd.) (I have grepped
> through their code. They do not have #ifdefs for NetBSD anywhere except
> for a library that I am not interested in, but they do have #ifdef FREEBSD
> at a few places. Besides having to explicitly provide -lpthread when
> linking etc is similar on NetBSD.)

There're other places where NetBSD differs from FreeBSD and Linux.
One of them is time_t which is wider than "usually." This explains why
you have XSB working on amd64, but it is only a hypothesis.

> With this change I could compile XSB on NetBSD 5.1 on amd64.
>
> However the same trick did not work for XSB on NetBSD 5.1 on i386. On
> i386, the C layer compiled fine, though a large part of XSB is written in
> Prolog itself and it gets compiled as a part of build where the basic xsb
> executable failed with memory violation message.
>
> I'm a long time user of XSB but with little knowledge of its internals.
> However I know one thing that they use pointers in some very peculiar ways
> to gain certain efficiency benefit. On "purist" architectures this could
> lead to problems. (Sorry, I do not know enough to say anything more
> specific than that.)

Can you point us to source tarball? Share changes you made?
Perhaps we could look at it.


-- 
HE CE3OH...



Home | Main Index | Thread Index | Old Index