Re: [Axiom-developer] gclcvs-2.7.0 and Axiom

[Stephen Wilson]

Camm Maguire <address@hidden> writes:

> Greetings!
> 
> Stephen Wilson <address@hidden> writes:
> 
> > Hello Camm,
> > 
> > I am interested in your opinion about the following error, which is
> > occurring for me during the compilation of src/algebra/tree.spad:
> > 
> > -------
> >  Error: 
> >  Fast links are on: do (si::use-fast-links nil) for debugging
> >  Signalled by |NEWLOOKUPINADDCHAIN|.
> >  Condition in |newLookupInAddChain| [or a callee]: 
> > INTERNAL-SIMPLE-TYPE-ERROR: #<FREE OBJECT 01ce05c8> is not of type 
> > FUNCTION: 
> > 
> >  Broken at |NEWLOOKUPINADDCHAIN|.  Type :H for Help.
> >   1 (Continue) Return to top level.
> >  BOOT>>:bt
> > 
> >  #0   newLookupInAddChain {loc0=|empty?|,loc1=((|Boolean|) $),loc2=#<vector 
> > 01f8d9f0>,loc3=#<vector 01f8d9...} [ihs=14]
> >  #1   genDeltaEntry {loc0=(|empty?| (#1=(|List| s) (|Boolean|) (|List| s)) 
> > (t (elt #1# #))),loc1=(el...} [ihs=13]
> >  #2   compForm3 {loc0=(|empty?| |ls|),loc1=|$EmptyMode|,loc2=(((# # # 
> > ...))),loc3=(((#1=# # #) (...} [ihs=12]
> >  #3   compForm {loc0=(|empty?| |ls|),loc1=|$EmptyMode|,loc2=(((# # # 
> > ...))),loc3=nil,loc4=speci...} [ihs=11]
> >  #4   setqSingle {loc0=#:g2300,loc1=(|empty?| 
> > |ls|),loc2=|NoValueMode|,loc3=(((# # # ...))),loc4=...} [ihs=10]
> >  #5   compSeq1 {loc0=((let #1=#:g2300 (|empty?| |ls|)) (|exit| 1 (if #1# # 
> > ...))),loc1=($),loc2...} [ihs=9]
> >  #6   compDefineCapsuleFunction {loc0=(def (|tree| |ls|) (nil nil) 
> > ...),loc1=|$EmptyMode|,loc2=(((# # # ...))),l...} [ihs=8]
> >  #7   compCapsuleInner {loc0=((mdef (|cycleTreeMax|) (nil) ...) (qsetrefv $ 
> > 7 ...) (|:| |t| $) ...),loc...} [ihs=7]
> >  #8   comp {loc0=(|where| (def (|Tree| s) (# nil) ...) (|:| s 
> > (|SetCategory|))),loc1=|$Empt...} [ihs=6]
> >  #9   SpadInterpretStream {loc0=1,loc1=(tim daly 
> > ?),loc2=t,loc3=nil,loc4="doc",loc5="spadhelp",loc6="axiom...} [ihs=3]
> >  #10   spad {loc0=|.axiom|,loc1=nil,loc2=nil,loc3=#<synonym stream to 
> > *terminal-io*>,loc4=25...} [ihs=2]
> > -------
> > 
> > The function |newLookupInAddChain| is defined in
> > src/interp/nrunfast.boot.pamphlet, and everything appears to be in
> > order on the Axiom side, however:
> > 
> > This is proving to be very difficult to debug, since I cannot
> > reproduce except by starting Axiom and immediately compiling the file.
> > If I do (si::use-fast-links nil) the problem goes away. Attempting to
> > trace any of the above functions, or even typing "1+1" at the axiom
> > prompt before compiling, the problem goes away.
> > 
> > This behavior is new, so I suspect a change in gclcvs. If this is a
> > GCL issue, I am hoping that you might be able to understand what this
> > problem could be, or perhaps suggest a technique to debug it.  If
> > there is anything I can do to provide more information please let me
> > know.
> > 
> 
> Unfortunately, this looks like a gc bug, which really cannot be found
> without running under gdb.  Last night I built silver with your patch
> and gclcvs compiled with --enable-debug  all the way up to the viewman
> stage (I didn't have libXpm.a installed) -- I take it this is past
> your problem area above.  Am retrying with more C optimization, but
> with -g in place of -fomit-frame-pointer.  Once I can reproduce, I'll
> post more.

Yes, the building of viewman is past the point of difficulty.

Appologies for passing around patches.  This is clearly not the most
efficient way to ensure our trees are identical.

I have uploaded a tarball of the current tree I am working with to
axiom-developer.org.  I believe you have an account there? It is
available at:

   /home/swilson/silver-gclcvs.tar.gz
 
Hopefully building off of this tree might yield the problem I am
seeing.  (As with my previous patches, you need gcl-2.7.0 ANSI
pre-built and in your path).

> I take it by 'new' you mean new vis a vis 2.6.8 -- or do you mean not
> present in a slightly older 2.7.0?  If so, what are the date bounds?

This is w.r.t 2.7.0.  I have not been following cvs as closely the
last few weeks as I have in the past but this change seems to have
occurred between the latest commits 2.7.0-78 and 2.7.0-79 (reading from
the debian changelog).


Thanks again for looking into this,
Steve

> 
> Take care,
> 
> > 
> > Many thanks,
> > Steve
> > 
> > 
> > 
> > 
> > _______________________________________________
> > Axiom-developer mailing list
> > address@hidden
> > http://lists.nongnu.org/mailman/listinfo/axiom-developer
> > 
> > 
> > 
> 
> -- 
> Camm Maguire                                          address@hidden
> ==========================================================================
> "The earth is but one country, and mankind its citizens."  --  Baha'u'llah