[Ecls-list] CVS back to normality

Juan Jose Garcia-Ripoll worm at arrakis.es
Tue Sep 17 08:20:05 UTC 2002


I solved the problems I caused because of uploading only a subset of the 
files, and which arised because of a bad synchronization between my laptop 
and the desktop computer.

Ok, after the excuses, a summary of what is new in the CVS tree.

* Errors fixed:

  - The arguments of GENSYM should not be remembered by ECLS. The
    current behaviour was wrong:
		> (gensym)
		> (gensym "F")
		> (gensym)

* System design:

  - Most symbols belonging to the LISP package and to the C core, are
    kept in a single array, "cl_symbols". The translator "dpp" and the
    routine SI::MANGLE-NAME, output the right name for any symbol
    which is in this array. This saves up to 20k in diferent platforms,
    speeds up initialization...

* Visible changes and ANSI compatibility:

  - ECL now uses a hashtable to store objects when *PRINT-CIRCLE* is
    on. This hashtable is not cleared when calling PRINT-OBJECT, so
    that circularities are also detected inside these methods.

  - *PRINT-CIRCLE* defaults to NIL.

* Known bugs:

  - The old mark&sweep garbage collector works no more. I long ago forgot
    to try building ECL with this garbage collector (i.e. with the
    --disable-boehm at configuration time), and right now it is loosing
    data somewhere. The code for this GC is extremely ugly and until I
    find time to find the leak, ECL can only be built with Boehm-Weiser
    GC (i.e. --disable-boehm is disabled :-)

It would be nice if people reading this list tried to compile ECL as it is. If 
I do not receive any negative feedback, I will be releasing a new package in 
one or two days.


http://www.arrakis.es/~worm		worm at arrakis.es
http://ecls.sf.net			yahoo-id: postdoc74

More information about the ecl-devel mailing list