[therion-users] Re: therion does a seg fault when I try and run it

[Wookey]

+++ Michael Lake [03-08-29 10:29 +1000]:
> Hi all,
> 
> > +++ Stacho Mudrak [03-08-28 09:14 +0200]:
> >>Uff - this is the first time therion does not work on linux!
> 
> :-)
> 
> Wookey wrote:
> > Mike is almost certainly the first person even to try a non-x86 version
> > (I haven't actually tried it on arm yet - I really must do and see if
> > similar problems arise). Is powerpc big or little-endian Mike
> 
> Hehe at me using a non-intel architecture. And I have some results for 
> you on Alpha LX164 architecture too :-)
> I'll comment on  that at the end of this email.
> Looking on google it seems PPC is big endian and Alpha is LE.(but i have 
> seem some comments that Alpha can emulate any endianness)

yes - Arm can do either endianness as well (and so can SH4) but the Debian 
version picks a
'standard' endianness for the architecture. I'm not sure which it is for
Alpha. ARM is LE in Debian.


> What suggestions now to debug this? I am keen to get this working on 
> both PPC and Alpha. 

Run it under strace to see where it actually dies. Look at the code giving
the char/unsigned char warnings - that's very likely to break something and
is suggestive of 'x86isms' in the code. Structure packing is different on
ARM from x86 and that can catch people out too, although I'd expect it to be
less of a problem with C++ than C. Maybe this is true on powerpc too?

It's also possible that the C++ compiler for powerpc has gremlins at the
moment - there has been a lot of transition recently from G++ 2.95 to 3, 3.1
3.2 and 3.3. Testing and unstable are now on g++3.3 but stable is still 2.95.

Certainly C++ compiling on ARM prior to about 3.2 was 'mostly broken' - I'd
expect powerpc to be better than that but compiler problems could easily
exist there too.

Wookey
-- 
Aleph One Ltd, Bottisham, CAMBRIDGE, CB5 9BA, UK  Tel +44 (0) 1223 811679
work: http://www.aleph1.co.uk/     play: http://www.chaos.org.uk/~wookey/