Re: [Axiom-developer] def.lisp

[Gabriel Dos Reis]

root <address@hidden> writes:

| > | In principle "user" can type-in parse tree by hand and call the 
translator.
| > | More realistically, one can add code which directly generates Spad like
| > | parse trees and then calls translator to output Aldor.
| 
| potentially aldor can also generate these parse trees.

Are you saying the Aldor compiler can parse SPAD but would not
itself translate the parse tree to Aldor?


Anyway, def.lisp was never removed from the actual compiler that is
installed.  Here is a documentation committed to build-improvements.

-- Gaby

2006-10-28  Gabriel Dos Reis  <address@hidden>

        * Makefile.pamphlet: Explain why def.$(OBJEXT) may be loaded
        twice. 

*** Makefile.pamphlet   (revision 15933)
--- Makefile.pamphlet   (local)
*************** The {\bf OPOBJS} list contains files fro
*** 305,310 ****
--- 305,320 ----
  ``old'' is something of a subtle concept as there were several
  generations of ``old'' and all meaning of the term is lost.
  
+ Notice that the object file [[def.$(OBJEXT)]] appears on both the
+ [[OPBJS]] and [[TRANOBJS]] lists.  In normal situation, parsing 
+ precedes translation; consequently the file [[def]] is loaded by the 
+ parser, so that it does not need to be reloaded by the translator. 
+ However, it may theoretically be that a translation could happen without 
+ prior parsing (in case someone types in a parse tree for SPAD code).  
+ Consequently, it must be ensured that [[def.]] is still loaded in that
+ configuration.  In the long term, the autoload machinery need
+ rethinking.
+ 
  <<environment>>=
  # These are autloaded old parser files
  OPOBJS=       ${AUTO}/parsing.$(OBJEXT)       ${AUTO}/bootlex.$(OBJEXT)       
\