[Gcl-devel] Re: more on Matt's bug report about pc-nqthm-1992, and *maybe* the compiler

[Camm Maguire]

Greetings!  I've checked this out a bit, and it seems as if we may
have run across another SGC garbage collection error, which have
proved exceedingly rare thus far, though I'm not 100% sure.  Just
remove the (si::sgc-on t) command in the save.lisp of the main
nqthm-1992 directory, or alternately fire up the image, do (si::sgc-on
nil), and save to a new version before compiling your pc-nqthm
version.  This works for me for now, but we definitely need to chase
down the apparent freeing of used objects in this example under sgc.

Take care,

Robert Boyer <address@hidden> writes:

> > It shows a call to compiler::analyze-regs that breaks.
> 
> I'll sure admit to being entirely at sea here, but my current best guess is
> that somehow the nqthm-1992 gcl image contains some corruption, caused
> perhaps in the save process, and where that corruption first causes enough of
> a problem to be detected is unrelated to its origin.
> 
> Bob
> 
> 
> 
> 

-- 
Camm Maguire                                            address@hidden
==========================================================================
"The earth is but one country, and mankind its citizens."  --  Baha'u'llah