Bill Allombert píše v Čt 19. 10. 2006 v 21:55 +0000: > > It should be compared to the buildlog on hobbes: > > <http://buildd.Debian.org/fetch.cgi?pkg=pari;ver=2.3.1-2;arch=m68k;stamp=1161007964> > > > > It took approximatively 24 hours (wallclock time) so it is approximately > > 3 times slower, but there is room for improvement. > > Actually that is not accurate: I build the full package while hobbes buit > only binary-arch, so aranym is not that slow. > > > Almost successfully because for some reason dpkg-shlibdeps failed to > > find most dependencies, so the package has missing dependencies. > > This was a general dpkg failure that was fixed by downgrading dpkg to > the Etch version (probably a FPU emulation bug).
Nothing is perfect, not even ARAnyM. I must admit that there was (yet another) MMU emulation error (pretty nasty to catch), that could decrement the stack pointer wrongly when an instruction storing something to stack was restarted because it tried to fetch its operand value from a different, unmapped memory page (yes, instruction on page boundary, again). A preliminary patch that fixes the issue for LINK instruction is in CVS but a more complete patch is expected in next release of ARAnyM. Sorry for that. Really. Petr P.S. detailed explanation of what I have found is in the ARAnyM-dev list archive but since the web interface reports an error in proxy configuration right now I can't provide you with a direct link to my post there. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
