ALERT: CLC-INTERCAL 1.-94.-2.2 escaped

[Claudio Calvelli]

CLC-INTERCAL has a new control structure, generalising both ABSTAIN FROM
and COME FROM! This is called "Roadworks". It allows to declare a section
of a program "Closed for roadworks", which makes it impossible to execute
it. Any attempts to do so will result in the program taking a "Diversion"
via a different segment. This new feature can also be referred to as
"Closure", which is particularly appropriate when one wants to confuse
programmers familiar with a different meaning of that word.

Unfortunately, in the process of testing this new feature, the secure
containment facility which prevents CLC-INTERCAL from heading out into the
world was temporarily closed off and the source code diverted via the
download server. In other words, it escaped again.

The escaped compiler can be found at:

https://uilebheist.srht.site/dist/CLC-INTERCAL-1.-94.-2.2/

Changes since 1.-94.-2.1 are detailed in the Changes files which can be found
all over the sources; or see the git log at:

https://git.sr.ht/~uilebheist/CLC-INTERCAL

As a very short summary, in alphabetic order:

* A microscopic change to sick.iacc results in an improved compiler
performance with no changes in the language accepted: for example,
the hello.i program (which is nothing short of cruelty to compilers)
now compiles approximately 15 times faster than before. Also, building
the whole compiler from sources takes a much shorter time.

* C-INTERCAL's ONCE and AGAIN statement modifiers are now supported; in
addition, they have a quantum version (ONCE AND AGAIN) which does not
exist in C-INTERCAL.

* Embedded INTERCAL in Perl sources. This used to be possible with 0.05 but
the functionality was temporarily removed in 1.-94.-8; it's now back,
so it's possible to have programs written partially in Perl and partially
in INTERCAL. Moreover, it is possible to specify different preloads in
the same source, so the INTERCAL parts could be a mix of CLC-INTERCAL,
C-INTERCAL or the original 1972. The possibilities for obfuscation are
endless.

* Implemented a mechanism by which a compiler (IACC source) can save
information about statements; this is used to implement the ONCE / AGAIN
feature, but also to count the ratio of statements containing PLEASE and
will eventually be used to produce an appropriate error for insufficient
or excessive use of PLEASE.

* Improved assignment to unary operators for bases > 4; it now works in a
lot more cases.

* Improved compatibility with C-INTERCAL's ABSTAIN FROM implementation:
when ABSTAINing FROM (or REINSTATing) a label, CLC-INTERCAL will check
if the corresponding statement is a GIVE UP and will transform the
ABSTAIN FROM / REINSTATE into a no-op if that's the case. C-INTERCAL
on the other hand executes it. This corresponds to two different
interpretations of the 1972 manual: both interpretations are valid.
When running in "imitate ick" mode, CLC-INTERCAL will now use the
C-INTERCAL interpretation, in all other modes it will use its own
interpretation, as documented since escape 0.01.

* Improved compatibility with C-INTERCAL's "select" binary operator. In
INTERCAL-72, the size of a number was determined at runtime, so that
the result of a select could be 16 or 32 bits depending on the actual
numbers it operated on. C-INTERCAL changed that to determine all sizes
at compile time, so that the size of a select is the size of its second
operand. CLC-INTERCAL kept the original "dynamic" specification. This
can produce different results if the select itself is the operand to an
unary operator. Starting from this escape, the size is determined at
runtime in most cases, but when compatibility with C-INTERCAL is desired
it is determined using the same compile time mechanism as they use.

* Improved compiler backend for object listing, which now includes more
information (see also the new backend below). Also permitted multiple
backends from the same session, for example ListObject,Run,DumpRegisters
would produce an object listing, then run the program, and finally
print the contents of all registers as they were at the end of execution.

* Made a start on a new optimiser, which for now does very little, but it's
a very effective little.

* Major Interpreter rewrite to remove "multi-function" entry points and
make them separate; this results in much easier to read code, AND
a slight performance improvement too!

* Modified some of the INET tests so that they work without a network.

* New compiler backend which dumps all register contents; it can be
called immediately after running a program to inspect the final state;
it will provide the information even if the program splatted, so that
could be used to try to figure out what happened.

* New test programs, both for the new features and additional tests for
existing features.

* Optimised version of ick's standard library and floating-point library:
this is bug-for-bug compatible with the INTERCAL sources, tested by
running a large number of library calls with random inputs and comparing
the outputs, but much faster. Unlike C-INTECAL's optimised library,
the "syslib" works in all supported arithmetic bases (floating-point
only work in base 2, like C-INTERCAL).

* Removed some obsolete opcodes which were reserved for the optimiser but
never used (the new optimiser works rather differently); also some other
opcodes which were undocumented and used internally have now disappeared,
but the new extension mechanism provides a documented way to do the same.

* Road closures / diversions, as mentioned above.

* Some bugs fixed, and to keep things balanced, new bugs added.

* Some statements and iasm (INTERCAL Assembler) opcodes have been renamed
to remove potentially offensive terms. The old names and syntax will
still work for compatibility with existing programs; however the
documentation will always prefer the new names.

* Support for C-INTERCAL's standard and floating-point library using the
same mechanism: if a program refers to labels belonging to these libraries
but does not define these labels, the compiler will try to include the
library. We do not distribute the library as there are questions about
its licensing. However we use it if it's available. Also, if the program
selects optimisations, it will be able to use the optimised version of
these libraries, (see above) which is distributed under the same licence
as CLC-INTERCAL itself and included in the distribution.

* Updated documentation, some of which even resembles reality.