Dear Maude Abusers,

A new alpha release can be accessed at SRI-CSL in:

~eker/public_html/Maude/Alpha86d/

or downloaded from:

http://www.csl.sri.com/~eker/Maude/Alpha86d/

Alpha release site authentication:

User Name: maudeabuser
Password: bughunter

This version is a Maude 2.2 release candidate and reports itself as
such. In addition to the normal Linux on i86 and Darwin on PowerPC
versions I have made binaries for:

Linux on x86-64		maude.linux64 (compiled with gcc 4.0.1)
Linux on PowerPC	maude.macLinux (compiled with gcc 4.0.1)
FreeBSD on i86		maude.freeBSD (compiled with gcc 4.0.3)

The x86-64 binary omits Doug Lea's malloc because of linking issues.
(The Darwin version has always omitted this version of malloc in favor
of Apple's version.)

I am considering dropping support for AIX (pain to build) and Alpha
(obsolete hardware). I'm not sure what to do about Solaris - it is a
pain to build (our system guys don't have a working gcc 4.0.1 on this
platform yet) - but Sun are still shipping SPARC-based workstations.
Does anyone care about this platform? 

Bug fixes
==========
(1) A stale pointer bug in the code handling invalidatation and
reevalutation of views. Found by Manolo Roldan and provoked by the
following code:

fth NAT* is
   protecting NAT .
   op * : -> Nat .
endfth

fmod NAT/{N :: NAT*} is
   sort Nat/{N} .
endfm

mod BUG is
  pr P{3} .
endm

view 3 from NAT* to NAT is
   op * to term 3 .
endv

mod P{P :: NAT*} is
  protecting NAT/{P} .
endm

fth NAT* is
   protecting NAT .
   op * : -> Nat .
endfth

fmod NAT/{N :: NAT*} is
   sort Nat/{N} .
endfm

mod BUG is
  pr P{3} .
endm

(2) fmod LIST-AND-SET in prelude.maude used inc rather than protecting
for its imports. Reported by Narciso.

(3) An uninitialized format attribute bug, visible under Purify.

(4) A memory leak where the module expressions for parameter theories
were not being deleted when a parameterized module is overwritten by a
newer module with the same name.

(5) A memory leak where identities for polymorphs were not being
deleted with the polymorph.

(6) A bug where identities for polymorphs were not being preprocessed;
provoked by the following example:

fmod NORM is
  sort Foo .
  ops 0 a : -> Foo .
  ops s f : Foo -> Foo .
  op _+_ : Foo Foo -> Foo [id: 0] .
  op _*_ : Universal Foo -> Universal [poly (1 0) right id: (s(0) + 0)] .
var X : Foo .
  eq a * s(X) = f(X) .
endfm

red a  .

Here _*_'s identity is not simplifed to s(0) before the equation is
semi-compiled and thus the match X |-> 0 is missed at runtime.

(7) A bug in the deletion of invalidated modules from a modules
metamodule cache.

(8) lesserSorts() is defined on types but only worked on sorts; this
is resolved by making it work on types. Reported by Narciso.

Other Changes
==============
(1) prelude.maude now contains QID-SET; META-MODULE now protects
  QID-SET * (op empty to none, op _,_ to _;_ [prec 43])
and has extra sorts NeSortSet, NeKindSet, NeTypeSet together with
extra subsort and operator declarations needed to preserve
preregularity and type associativity of _;_. This change was requested
by Carolyn.

(2) sortLeq() is now defined and implemented over types rather than
sorts for symmetry with the lesserSorts() fix.

Steven