- Added TERMLISTERROR constant, and corresponding tests.
Note that this will not work in many contexts, because only NULL is
usually considered to be a special value. It is purely intended for
the new type evaluation functions in type.c.
terms were destroyed before it could be tested whether they were in
sys->variables. Thus, garbage was left in sys->variables. This has
gone undetected because it was never really used. Hmpf.
Note that for variables that are not instantiated, only the variable
and the type info is shown. For instantiated variables, both are
shown.
In this output the attribute 'free' should be ignored, as its
output is not accurate here.
--tupling=n * 0: right-associative; 1: left-associative; others are
reserved for future use.
--ra-tupling * Sets the default, but is there for symmetry.
instead of the default right-associative tupling. Note that this only
matters for full typeflaw matching.
- Adapted multi-nsl test script to test for both association variants.
code was added (two weeks ago), bindings were changed without changing
the state of the graph closure buffer. This resulted in possible
missed loops: thus, reports of broken claims could be found in the
output, in inconsistent states. (Reported bij Gijs Hollestelle.)
Note that this only influences the current development release, and
not any previous results.
the (monstrously large) system structure, there is now a global
'switchdata' structure originating in switches.c. This makes it much
easier to see what's happening.
* Note: although this code has been tested, there might be some
hiccups, because doing multiple search&replace actions over all
files is bound to cause some problems.
* removed <term> wrappers
* added <const> wrappers
* removed <role><term> construct, now <rolename>R</rolename>
constructs.
* added <variables> section.
* variable substitutions are followed through in runs. Thus, only
unbound variables occur in the semitrace.
* added the untested claims back in, so that all events in a
role/semitrace are now shown. Note that they can be disabled
again by using the new '-H' switch.
expanded explicitly. This solves a long-standing issue with {k}k
decryption to yield k. Needs some testing to ensure that it did not
introduce any new errors.
