Dear Maude Abusers,

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

~eker/public_html/Maude/Alpha90a/

or downloaded from:

http://www.csl.sri.com/~eker/Maude/Alpha90a/

Alpha release site authentication:

User Name: maudeabuser
Password: bughunter

The version reports itself as Maude 2.4 and will be release as such to
coincide with WRLA 2008 unless show stoppers are reported.

Changes
--------
(1) The GNU libsigsegv library is used to distinguigh between true
segmentation faults and stack overflows and so stack overflows are now
reported with an informative message. Previous behavior was to catch
true segmentation faults but not install an alternative signal stack
so stack overflows would go uncaught, with the OS terminating the
process. This would often confuse users. This library contains low
level platform specific code to heuristically distinguish between the
two conditions for many platforms. It has been added to the list of
prerequisites for building Maude from source.

(2) The semi-compilation of huge righthand sides has been optimized.
Consider an equation of the form

eq lhs = l(a1, l(a2, l(a3, ... l(a10000, a10001) ... ))) .

Here there were 3 issues that caused quadratic semi-compilation time
due to the optimization of slots in the substitution for storing
partly built instances.
(i) Left to right compilation in this case produced quadratically many
conflicts between slots. This is now resolved by compiled largest
subterms first.
(ii) The conflict detect used to compare the lifetime of every slot to
every other slot to detect conflicts. Now a shortlist of potential
conflict slots is maintained.
(iii) The graph colorer used an adjacency matrix. It now uses
adjacency sets.
These changes mean that examples such as the one above can be
semi-compiled in linear rather than quadratic time. This was motivated
by a pathological example from Manuel, that had 20001 subterms needing
to be assigned slots.

(3) The construction of free theory discrimination nets has been
optimized. Although is it still worst case exponential, a number of
subalgorithms which were best case O(n^2) are now O(n log n) where n
is the number of equations with a given top symbol. These
optimizations were motivated by a pathological from Manuel that had
10000 equations for one symbol.

(4) A lot of redundant code and some redundant classes are deleted.

Steven