Ann: SWI-Prolog 7.1.11

Hi,

I've uploaded SWI-Prolog 7.1.11. The bad news is that there is little
new. The good news is that appearently you all did a good deal of
hacking and reported many issues with or without fix :-)  Highlights:

   * print/1 has been modified as discussed, i.e., the default is
     now quoted write and there is a prolog flag to control print/1.
     The flags toplevel_print_options and debugger_print_options have
     been renamed to get more consistent naming.

   * Fixed several possible crashes and memory leaks (mostly Keri
     Harris).

   * Two fixes to the RDF store that can cause crashes.  Also updated
     ntriples parser for RDF 1.1.

   * Various fixes to the development environment, notable to the
     handling of name() and DCGs.

And a lot of small stuff. See below. Also cherry-picked many of these
patches to the stable git version. If nothing bad shows up with any of
these the next week, I'll release stable 6.6.5.

	Enjoy --- Jan


=========================================
SWI-Prolog Changelog since version 7.1.10
=========================================

  * FIXED: off-by-one error in signal handlers

  * ENHANCED: Avoid tracing into arithmetic expansion

  * MODIFIED: print/1 now used quoted output by default.  Its behaviour
    is now defined by the new Prolog flag =print_write_options=.

  * MODIFIED: Renamed flag =toplevel_print_options= into
    =answer_write_options=

  * MODIFIED: flag =debugger_print_options= is renamed into
    =debugger_write_options= for consistency with =print_write_options=.

  * CLEANUP: Use foreign_t rather than word for write predicate
    implementations.

  * FIXED: Possible non-termination in code walker due to incomplete
    source information.

  * FIXED: Development tools for handling name() at various places.

  * DOC: Added det, nondet, semidet and choice point to the glossary.

  * FIXED: Feedback output of commandline tracer in Windows swipl-win.exe
    was broken.  Grishma Jena.

  * FIXED: check_installation/0 for library(pce).

  * FIXED: prolog_trace_interception/4 on unify caused attribute wakeups
    not to be called.  Edison Mera.

  * FIXED: '$find_predicates/2' should not return system: for things
    imported from the library.

  * FIXED: Some enhancements to listing/1.  Notably make listing(module:_)
    list all predicates in module.  Jan Burse.

  * FIXED: Possible memory leak in sub_atom/5 and sub_string/5.
    Keri Harris.

  * FIXED: use of uninitialized memory in peek_string/3

  * FIXED: thread_create_in_pool/4: generate exception if new pool creation
    hook fails.

  * FIXED: Also need to check we have a thread in the libunwind version.
    Keri Harris.

  * FIXED: Missing break; in ar_pow, leading to an assertion error.
    Markus Triska.

  * ENHANCED: CLP(FD): Stronger exponentiation.  Previously:

       %?- A^2 #= C, C in 253..257, A #> 0.
       %@ A in 15..16,
       %@ A^2#=C,
       %@ C in 253..256.

    Now:

       %?- A^2 #= C, C in 253..257, A #> 0.
       %@ A = 16,
       %@ C = 256.

  * FIXED: save_backtrace() if threads are not yet initialized.
Keri Harris.

  * ADDED: hook thread_pool:create_pool/1 to allow for lazy creation of
thread pools.

  * FIXED: Illegal casts in C++ between pointers.  This patch complements
a patch to the cpp package.

  * ADDED: parse_time/3: support numeric RFC 1123 timezones RFC 1123
    section 5.2.14 says, "There is a strong trend towards the use of
    numeric timezone indicators, and implementations SHOULD use numeric
    timezones instead of timezone names."  Since the RFC recommends
    numeric timezone support, parse_time/3 ought to support them.

  * FIXED: Integer size issues with creating saved states.  Roberto
    Bagnara.

  * FIXED: Max stack-size warning for 64-bit platforms

  * ENHANCED: safe_goal/1: Do not try to specify for a more specific goal
    if the generic goal is the same as the specific one.

  * FIXED: Possibly listing of predicates multiple times due to changed
    semantics of list_to_set/2.

  * FIXED: Handling of {G} in //-annotated arguments.  Samer Abdallah



===============
Package archive
===============

  * PORT: Omit filter property for older libarchive versions.
    Carlo Cappelli.

============
Package clib
============

  * FIXED: Allow for tcp_bind(localhost:Port), i.e., bind to an interface,
    but leave the port to the system.

  * DOC: Updated tcp_bind/2 docs


===========
Package cpp
===========

  * FIXED: Illegal cast of C++ pointers Hello SWI-Prolog developers,

    and thank you for your work!

    Here is a patch proposal about the CPP package that involves also
    modifications in src/SWI-Prolog.h (in a separate mail).

    8<--------------------------------------------------------------

    ISO-C++ forbids casting pointers-to-functions to pointers-to-objects.
    This patch fixes some casts that issue a warning in GCC.

============
Package http
============

  * TEST: Added test to verify that the server closes the connection if
    requested to by the client.

  * FIXED: Make sure http_server(http_dispatch, localhost:Port) is
    documented and works.

  * FIXED: Bug#172: connection(close) is ignored for requests that are
    handled through exceptions.  This results in waiting for the server
    keep-alive timeout for clients that assume that the server will close
    the connection.  Raivo Laanemets.

  * DOC: Make library linkable.

  * FIXED: Bug#170: http_open/3 with empty server response causes confusing
    error message.  Raivo Laanemets

  * FIXED: Bug#168: json_read/2 can fail on illegal input.  Should be
    det and throw an exception.  Raivo Laanemets.

  * FIXED: Conditionally load code that depends on is_dict/1

============
Package odbc
============

  * FIXED: possible memory leak when reading large strings from database

  * FIXED: possible crash when reading large strings from database


================
Package pengines
================

  * ENHANCED: Use lazy thread-pool creation provided by latest SWI-Prolog.

  * DELETED: pengine_assert/1 and friends

  * ADDED: plans

  * ADDED: list of remaining bugs

  * ADDED: pengine_assert/1 and friends

  * MODIFIED: current_pengine/4 -> current_pengine/5

  * CLEANUP: Some layout, white space errors and missing declarations
    for new arguments.


==============
Package plunit
==============

  * ENHANCED: Avoid trapping the tracer on the test internals.

==============
Package semweb
==============

  * FIXED: Use defferred finalization to get rid of predicate clouds.
    Fixes an early free when closing a query.

  * TEST: Start random tests from printed seed, so we can replay them.

  * TEST: Fixed replaying of failed rdfs:subPropertyOf sequences

  * FIXED: Literal reference count was 24 bits and not checked for
    overflow.  Now 32-bit with assert-based overflow checking.

  * FIXED: We must always lock queries.write.lock when removing a triple
    in GC, both to update the hash chains and to ensure consistent update
    of triple->linked.

  * FIXED: Use atomic inc/dec to make sure the count for the number of
    triples per graph is accurate.

  * FIXED: rdf_load/2: Actually return the loaded graph when using
    graph(Var)

  * FIXED: rdf_load/2 must synchronize concurrent loads on the source
    rather than the graph to facilitate loading quad/trig files.

  * ADDED: Media types for ntriples and nquads

  * OOPS: Revert "CLEANUP: library(rdf_ntriples): synchronise iriref
    reading" This reverts commit c58c0d2ff233e41e478dfbfe11365507ec221d9a.
    I mistakingly though that the iriref was synchronised between Turtle
    and ntriples.  This does not seem to be the case.

  * CLEANUP: library(rdf_ntriples): synchronise iriref reading

  * MODIFIED: library(rdf_ntriples): synced character classification
    with Turtle.  Effectively makes detection of language tags and blank
    nodes compatible with the RDF 1.1 spec.

===========
Package ssl
===========

  * FIXED: memory leak when ssl_open/3 fails

=================
Package swipl-win
=================

  * BUILD: Avoid including packs in the app.

============
Package xpce
============

  * ENHANCED: Avoid a non-terminating loop of errors if (some component
    of) xpce cannot be found.  It is very hard to recover cleanly from
    this condition, so we just terminate the process :-(