Fixes <http://bugs.gnu.org/26261>.
Reported by Zefram <zefram@fysh.org>.
* module/srfi/srfi-19.scm ("define directives"): N padding increased from 7 to 9
* test-suite/tests/srfi-19.test ("date->string"): New test.
* module/srfi/srfi-18.scm (mutex-unlock!): Implement the
ignore-unlock-errors and wait-condition-variable behavior of
mutex-unlock! directly, without relying on Guile.
* module/srfi/srfi-18.scm (<mutex>, with-thread-mutex-cleanup)
(make-mutex, mutex-state, abandon-mutex!, mutex-lock!): Manage
"abandoned" bit on Scheme side with no need for thread cleanup
handler.
* module/srfi/srfi-18.scm (<condition-variable>): New data type.
(make-thread): Use srfi-18 interfaces.
(mutex-unlock!): Adapt to optional cond argument being disjoint from
Guile condition variables.
(make-condition-variable, condition-variable-signal!)
(condition-variable-broadcast!): Adapt.
* module/srfi/srfi-18.scm (<mutex>): Define as a wrapper type around
Guile mutexes.
(thread-mutexes): New variable.
(with-thread-mutex-cleanup): New facility to abandon mutexes on
SRFI-18 thread exit. Not yet used.
(make-thread): Use SRFI-18 interfaces.
(make-mutex): Reimplement for our boxed mutexes.
(mutex-state): Adapt.
(mutex-lock!): Adapt.
(mutex-unlock!): Adapt.
* test-suite/tests/srfi-18.test: Don't assume that SRFI-18 mutexes are
the same as Guile mutexes.
* module/srfi/srfi-18.scm (%cancel-sentinel, thread-terminate!): Just
use cancel-thread to cause the thread to return a sentinel value.
(%timeout-sentinel): Rename from %sentinel.
(thread-join!): Adapt and transform %cancel-sentinel to a
&terminated-thread-exception.
* module/srfi/srfi-18.scm (make-thread): Inline some helpers, and use
just one catch block.
(thread->exception): Move up definition.
(exception-handler-for-foreign-threads): Use this as the default
handler, not the one that squirrels away exceptions in
thread->exception.
(thread-terminate!): Don't instate an exception handler for the thread
cleanup proc.
* module/srfi/srfi-18.scm (with-exception-handlers-here): New function.
(wrap): Remove.
(thread-join!, mutex-lock!): Use with-exception-handlers-here instead
of the call/cc+wrap mess.
* module/srfi/srfi-18.scm: Use srfi-35 conditions instead of our
home-grown equivalent system.
(thread-exception-handlers): Remove unused table.
(srfi-18-exception-handler): Always capture key consed to args; no
special case for bare key.
* test-suite/tests/srfi-18.test (provided?): Adapt to reason always
being key+args.
* module/srfi/srfi-18.scm: Use srfi-34 internally with srfi-34: prefix.
(current-exception-handler): Be a parameter.
(with-exception-handler): Adapt to current-exception-handler change.
(thread-join!, mutex-lock!): Adapt to use srfi-34: prefix.
* module/ice-9/boot-9.scm (map): Rewrite to be recursive and pure
instead of iterative and effectful. At best this is faster; at worst
it is slower. In any case it resolves continuation-related issues.
* module/srfi/srfi-1.scm (fold): Specialize the two-arg case.
(map): Rewrite to be recursive.
* test-suite/tests/r5rs_pitfall.test (8.3): Update for new expected map
behavior.
* module/ice-9/match.upstream.scm (match-next): Inline a call to
"error", so the new CSE pass will see this case as a bailout.
* module/srfi/srfi-9.scm (throw-bad-struct): Reimplement as a syntax
rule, so that the CSE pass sees the "throw" call.
* module/ice-9/boot-9.scm (map):
* module/srfi/srfi-1.scm (map): Simplify the implementations to check
for list? beforehand. It's faster, and it will be needed if we decide
to go recursive.
* module/ice-9/boot-9.scm (for-each):
* module/srfi/srfi-1.scm (for-each): Re-implement one-list case using an
explicit check for list? instead of the tortoise-hare thing. Seems to
be faster!
* module/srfi/srfi-4.scm (define-bytevector-type): For the predicates
and length accessors, only accept bytevectors. Since arrays don't
work for u32vector-ref et al, they shouldn't pass u32vector?.