14.23.3 Bugs Fixed

• Do-loops (do/2): documentation clarified.
• During compilation with the debugger switched on, it would spuriously stop at some structs predicates.
• Documentation incorrectly mentioned non-ISO float syntax sign inf and sign nan.
• Error handling of mod/2 and friends.
• Incorrect result when using compare/3 et al. on atoms and functors with NUL characters in their names, e.g. 'foo\0\bar'. NUL is now treated as smaller than all other character codes.
• Attributed variables that have been deterministically bound were not garbage collected.
• Compiler warnings referring to variable names were ugly.
• use_module/2 et al. with empty import list no longer generate a message about “module ... imported into ...”.
• Efficiency bug in SP_query() and friends.
• Decompiled code sometimes missed var/1 and nonvar/1 tests.
• Efficiency bug affecting call_cleanup/2, findall/3 and friends, and I/O.
• Minor issue when writing terms if '.' or '$VAR' has been declared as operators.
• call/1 at al. could hang, uninterruptibly, for certain cyclic goals.
• SP_put_float() will now fail if the argument is not finite (i.e. if the argument is +/-inf or NaN).
• Error handling of bb_get/2, bb_put/2, bb_delete/2, bb_update/3.
• On OS X, sicstus would sometimes fail to set the locale, and would fail to launch. This could prevent M-x run-prolog from starting Prolog in Emacs.
• library(atts): documentation and error handling.
• ugraphs:transitive_closure/2 and wgraphs:transitive_closure/2: wrong answer for cyclic graphs.
• lists:same_length/2, lists:same_length/3, lists:rotate_list/2 and lists:rotate_list/3 left a choicepoint behind for same cases when an argument was a partial list.
• ugraph:random_ugraph/3 would loop if its output argument failed to unify with the resulting graph.
• ordsets:ord_intersection/4 left a choicepoint behind.
• In the Emacs mode, some code formattng variables have been marked as “safe local variables” so that they can be set when opening a file, without eliciting a prompt from Emacs.
• CLPFD:
  • - Finite but very large domains could crash.
  • - Some propagators could access uninitialized data.
  • - Wrong attribute projection of minimum/2 and maximum/2.