summaryrefslogtreecommitdiffstats
Commit message (Collapse)AuthorAgeFilesLines
...
* Missing notes about global lexical variables.Kaz Kylheku2015-08-201-8/+43
| | | | | | * txr.1: Note that TXR Lisp has global lexicals, and that some library variables are examples of such, and that defparm is defparameter in CL.
* Wording in description of quasiquote.Kaz Kylheku2015-08-201-2/+2
| | | | | * txr.1: An unquote's evaluation should be contrasted with the quasiquote's suppression of it, not with the quote.
* Version 113.txr-113Kaz Kylheku2015-08-144-4/+44
| | | | | | | | * RELNOTES: Updated. * configure, txr.1: Bumped version and date. * share/txr/stdlib/ver.tl: Likewise.
* Remove useless member from parser structure.Kaz Kylheku2015-08-192-3/+3
| | | | | | | * parser.c (parser_common_init): Use local variable to capture output of yylex_init, rather than yyscan structure member. * parser.h (struct parser): Member yyscan removed.
* Fix parser bug caused by wrong kind of weak hash table.Kaz Kylheku2015-08-191-1/+1
| | | | | | | | | | | | | | This bug causes the parser associated with a stream to suddenly disappear while reading forms from the stream. Parsing continues, but with a new parser which does not carry the lookahead token from the previous parse. So for instance, we miss the opening parenthesis of the next form. * parser.c (parse_init): The stream_parser_hash must be a hash table with weak keys, but not weak values. We want the association to go away only if the stream becomes unreachable, not if the parser becomes unreachable while the stream is still reachable.
* Provide conservative marking interface.Kaz Kylheku2015-08-193-29/+28
| | | | | | | | | | | | | | | | | | | GC now exports a function for marking something that might not be an object, instead of a function for testing. The previous way wasn't integrated with Valgrind properly, and didn't observe the FREE flag. * gc.c (gc_is_heap_obj): Function removed. (mark_obj_maybe): New static function, with body consisting of code moved from mark_mem_region. (mark_mem_region): Moved code replaced by call to mark_obj_maybe. (gc_conservative_mark): New function, wraps mark_obj_maybe. * gc.h (gc_conservative_mark): Declared. (gc_is_heap_obj): Declaration removed. * parser.c (yy_tok_mark): Use gc_conservative_mark instead of gc_is_heap_obj check and gc_mark.
* Fix broken @@@<n>/@@@rest references in quasiliterals.Kaz Kylheku2015-08-192-2/+6
| | | | | | | | * parser.y (quasi_meta_helper): When obj is a sys:var, leave it alone; don't add another layer of var. Also, do the same if it is a sys:expr. * tests/012/quasi.tl: Added test case.
* Quasiquote regression from 110.Kaz Kylheku2015-08-192-5/+13
| | | | | | | | | | | | | | | | The problem is that one-argument function calls like @(whatever arg) in a quasiliteral being turned into sys:var items. * parser.y (quasi_meta_helper): Remove bogus check on length. The default case is now var, so the var_s check actually matters. The integerp check for the argument of a var form didn't do anything because the entire if statment conditionally selected a useless goto. Removing it for consistent treatment of var items. * tests/012/quasi.tl: Some new test cases involving @rest. These new tests pass whether or not we have that integerp(second(obj)) test in the quasi_meta_helper function. Either way @rest and @@rest produce the same thing.
* Troff formatting macros must not generate control sequences.Kaz Kylheku2015-08-191-4/+4
| | | | | | | | Reported by Dave Love. * txr.1 (cod1, cod2, meti, TP*): When inserting material at the start of a line, add the \& zero width character, so the line isn't interpreted as a control sequence.
* Fix man page markup errors.Dave Love2015-08-171-22/+22
| | | | | * txr.1: Numerous misspellings of macro names are repaired.
* Better error reporting for some indirect calls.Kaz Kylheku2015-08-181-18/+24
| | | | | | | * lib.c (callerror): New static function. (generic_funcall): Use callerror to report errors against the function rather than the uninformative call: prefix. (wrongargs): Implement in terms of callerror.
* Bugfix: forgotten zaps could lead to spurious retention.Kaz Kylheku2015-08-171-3/+3
| | | | | * lib.c (generic_funcall): Added forgotten z() calls to zap arguments in some variadic cases.
* Streamlining format with new args module.Kaz Kylheku2015-08-171-66/+48
| | | | | | | | | | | * stream.c (vformat): split into a new function called aformat, and a small wrapper which retains the vformat name. (aformat): New function formed from most of vformat. Takes struct args * instead of va_list. (formatv): Big switch statement hack gone. Now the function fills in a struct args, allocated on the stack, with an accurate size based on the args list. This is passed directly to aformat.
* Start of new way of handling variable arguments.Kaz Kylheku2015-08-173-1/+162
| | | | | | | | * Makefile (OBJS): Add args.o object. * args.c: New file. * args.h: New file.
* Improved wording about auto-loading.Kaz Kylheku2015-08-161-5/+6
| | | | | * txr.1: In the stdlib variable description, rewrote the text describing the macro-time autoloading scenario.
* Detect alloca.Kaz Kylheku2015-08-151-0/+27
|
* Minor dependency fix in Makefile.Kaz Kylheku2015-08-151-3/+6
| | | | | | * Makefile (y.tab.c, y.tab.h, lex.yy.c): Express dependency of these on the configuration material via the DEP macro rather than directly. That way ABBREV can reflect over the dependency.
* Whitespace.Kaz Kylheku2015-08-151-1/+1
|
* Remove include guards from MPI headers.Kaz Kylheku2015-08-152-12/+0
| | | | | * mpi/mpi-config.h, mpi/mpi.h: These headers do not require include guards. They included only once.
* Allow slashes in regex passed to regex-parse.Kaz Kylheku2015-08-152-19/+47
| | | | | | | | | | | * parser.l (SREGEX): New start state, for stand-alone regex parsing. (grammar): All REGEX state rules are active in the SREGEX state also. The rule for the / character returns a REGCHAR if in the SREGEX state, so it is treated as an ordinary character. * txr.1: Updated regex-parse documentation about the treatment of the slash. Also added notes about double escaping when a string literal is passed to regex-parse.
* Regenerate parser and scanner if config changes.Kaz Kylheku2015-08-151-3/+3
| | | | | | * Makefile (lex.yy.c, y.tab.c, y.tab.h): Add missing dependencies on config.h and config.make so that these files are regenerated if we touch the configuration (for instance change the YACC variable).
* Get Berkeley Yacc port of the parser working again.Kaz Kylheku2015-08-141-0/+7
| | | | | | | | | | | | | * parser.y (byacc_fool): New grammar nonterminal symbol and dummy rule set. (spec): Use dummy byacc_fool to create a fake continuation in the grammar, so the Berkeley-Yacc-generated parser doesn't throw a syntax error. Our YYACCEPT prevents the byacc_fool part from consuming more than one token of lookahead. Bison doesn't need this because it has $default actions which reduce regardless of the lookahead token. BYacc insists on reducing only if it can match $end (end of input), and not other tokens, which constitute syntax errors.
* Formatting fix under clamp function.Kaz Kylheku2015-08-141-1/+1
| | | | * txr.1: .meta -> .metn
* Version 112.txr-112Kaz Kylheku2015-08-146-370/+423
| | | | | | | | | | * RELNOTES: Updated. * configure, txr.1: Bumped version and date. * share/txr/stdlib/ver.tl: Likewise. * tl.vim, txr.vim: Regenerated.
* Fix TEXT macro clash on Windows.Kaz Kylheku2015-08-141-0/+1
| | | | | * txr.c (TEXT): Undefined this macro after including <windows.h>. It clashes with a TEXT macro in y.tab.h.
* Remove unwanted yyparse declaration from y.tab.h.Kaz Kylheku2015-08-142-1/+9
| | | | | | | | | | * Makefile (y.tab.c): Putting in an ugly workaround for an obnoxious new behavior introduced in Bison 3.x, which breaks our build on platforms that have a newer Bison. After generating y.tab.h, we remove the unwanted declaration with sed. * parser.y (yyparse): Declare, since y.tab.h doesn't any more, and the newer Bison's parse skeletons expect it.
* Word splices not quite on board with consing dot handling.Kaz Kylheku2015-08-141-2/+2
| | | | | | | * parser.y (r_exprs): The WSPLICE and QWSPLICE syntax at the front of a list must now initialize the terminating atom to unique_s, not to nil. Without this we get mysterious "misplaced consing dot" errors (even though no consing dot occurs).
* New function, clamp.Kaz Kylheku2015-08-134-0/+45
| | | | | | | | | | * eval.c (eval_init): Register clamp as intrinsic function. * lib.c (clamp): New function. * lib.h (clamp): Declared. * txr.1: Documented.
* Observe display width in format.Kaz Kylheku2015-08-132-9/+59
| | | | | | | | | | | | | Take into account string display width in field trimming and padding calculations, including situations where only half a character fits. * stream.c (calc_fitlen): New function. (vformat_str): Revised to use calc_fitlen. If calc_fitlen indicates that there is no trimming or padding, then use put_string rather than put_char. * txr.1: Update description of format with regard to use of dispaly width in field calculations.
* Print object readably in class mismatch error.Kaz Kylheku2015-08-131-1/+1
| | | | * lib.c (class_check): Use ~s to print offending object.
* New display-width function.Kaz Kylheku2015-08-134-0/+51
| | | | | | | | | | * eval.c (eval_init): Register display-width intrinsic. * lib.c (display_width): New function. * lib.h (display_width): Declared. * txr.1: Documented display-width.
* Floating-point constant tightening.Kaz Kylheku2015-08-122-11/+12
| | | | | | | | | | | | | * parser.l (grammar): Change order of rule which recognizes FLODOT with a one-character trailing context other than a dot, and the rule which diagnoses trailing junk. The issue is that this order gives the wrong interpretation to 123.E, treating it as 123. followed by E rather than trailing junk, like in the case of 123.0E or 123.B. * txr.1: Adding the valid example 1.E5. Removing references to dot as consing dot. Fixed documentation which says that 1.E is 1 followed by a consing dot and E. The wrong behavior in fact produced 1.0 followed by E. No consing dot semantics.
* Improvement in hash bang code.Kaz Kylheku2015-08-121-6/+3
| | | | | * parser.c (read_eval_stream): Simplify hash bang code and avoid creating a string that might not end up being used.
* Use new pushback token priming for single regex parse.Kaz Kylheku2015-08-124-21/+38
| | | | | | | | | | | | | | | | | | | | | | | * parser.h (enum prime_parser): New enum. (prime_parser, prime_scanner, parse): Declarations updated with new argument. * parser.c (prime_parser): New argument of enum prime_parser type Select appropriate secret token for regex and Lisp case. Pass prime selector down to prime_scanner. (regex_parse): Do not prepend secret escape to string. Do not use parse_once function; instead do the parser init and cleanup here and use the parse function. (lisp_parse): Pass new argument to parse, configuring the parser to be primed for Lisp parsing. * parser.l (grammar): Rule producing SECRET_ESCAPE_R removed. (prime_scanner): New argument. Pop the scanner state down to INITIAL. Then unconditionally switch to appopriate state based on priming configuration. * parser.y (parse): New argument for priming selection, passed down to prime parser.
* Fix regression in string field formatting.Kaz Kylheku2015-08-121-4/+0
| | | | | * stream.c (vformat_str): Remove stray bogus code that unconditionally outputs extra left padding.
* Revision to .. syntax.Kaz Kylheku2015-08-122-31/+18
| | | | | | | | | | | | | * parser.y (r_exprs, n_expr): Move the DOTDOT syntactic sugar rule from r_exprs to n_expr, where it is much simpler. This also means that the a..b syntax is now an expression by itself; it need not be enclosed in a list. The DOTDOT operator is made right associative; or rather its existing %right declaration is now activated. * txr.1: Remove documentation stating that the .. notation must be used in a list, and not in the dotted position of an improper list. Document the behavior in the dotted position, and document right associativity.
* Crafting a better parser-priming hack.Kaz Kylheku2015-08-129-41/+97
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | The method of inserting a character sequence which generates a SECRET_TOKEN_E token is being replaced with a purely token based method. Because we don't manipulate the input stream, the lexer is not involved. We don't have to flush its state and deal with the carry-over of the yy_hold_char. This comes about because recent changes expose a weakness in the old scheme. Now that a top-level expression can have the form expr.expr, it means that the Yacc parser reads one token ahead, to see whether there is a dot or something else. This lookahead token is discarded. We must re-create it when we call yyparse again. This re-creation is done by creating a custom yylex function, which can maintain pushback tokens. We can prime this array of pushback tokens to generate the SECRET_TOKEN_E, as well as to re-inject the lookahead symbol that was thrown away by the previous yyparse. To know which lookahead symbol to re-inject is simple: the scanner just keeps a copy of the most recent token that it returns to the parser. When the parser returns, that token must be the lookahead one. The tokens we keep now in the parser structure are subject to garbage collection, and so we must mark them. Since the YYSTYPE union has no type field, a new API is opened up into the garbage collector to help implement a conservative GC technique. * gc.c (gc_is_heap_obj): New function. * gc.h (gc_is_heap_obj): Declared. * match.c: Include y.tab.h. This is now needed by any module that needs to instantiate a parser_t structure, because members of type YYSTYPE occur in the structure. (parser.h can still be included without y.tab.h, but only an incomplete declaration for the parser strucure is then given, and a few functions are not declared.) * parser.c (yy_tok_mark): New static function. (parser_mark): Mark the recent token and the pushback tokens. (parser_common_init): Initialize the recent token, the pushback tokens, and the pushback stack index. (pushback_token): New static function. (prime_parser): hold_byte argument removed. Body considerably simplified. The catenated stream trick is no longer required. All we do here is set up two pushback tokens and prime the scanner, if necessary, so it is in the right start state for Lisp. * parser.l (YY_DECL): Take over definition of scanning function, renaming to yylex_impl, so we can implement yylex. (grammar): Rule which produces SECRET_ESCAPE_E token removed. (reset_scanner): Function removed. (yylex): New function. * parser.h (struct parser): Now only forward-declared unless y.tab.h has been included. New members, recent_tok, tok_pushback and tok_idx. (yyset_hold_char): Declared. (reset_scanner): Declaration removed. (yylex): Declared (if y.tab.h included). (prime_parser): Declaration updated. (prime_scanner): Declared. * Makefile: express new dependency on existence of y.tab.h of txr.o, match.o and parser.o.
* Catenated stream reads must not close last stream.Kaz Kylheku2015-08-121-6/+9
| | | | | | * stream.c (cat_get_line, cat_get_char, cat_get_byte) Only close the head stream when popping it off the list.
* Diagnose ambiguous floats like (a b).4 and x.y.5Kaz Kylheku2015-08-101-0/+30
| | | | | | | | These look like integers involved in qref dot syntax. * parser.l (DOTFLO): New pattern definition. (grammar): New rules for detecting cramped floating literals.
* Dot with no whitespace generates qref syntax.Kaz Kylheku2015-08-105-6/+71
| | | | | | | | | | | | | | | | | | | | | | | a.b.(expr ...).c -> (qref a b (expr ...) c) Consing dot requires whitespace. * eval.c (qref_s): New symbol global variable. (eval_init): Initialize qref_s. * eval.h (qref_s): Declared. * parser.l (REQWS): New pattern definition, required whitespace. (grammar): New rules to scan CONSDOT (space required on both sides) and LAMBDOT (space required after). * parser.y (CONSDOT, LAMBDOT): New token types. (list): (. n_expr) rule replaced with LAMBDOT and CONSDOT. (r_exprs): r_exprs . n_expr consing dot rule replaced with CONSDOT. (n_expr): New n_expr . n_expr rule introduced here for producing qref expressions. (yybadtoken): Handle CONSDOT and LAMBDOT. * txr.1: Documented qref dot.
* Count East Asian Wide and Full Fidth chars as two columns.Kaz Kylheku2015-08-104-7/+74
| | | | | | | | | | | | | | | | | * regex.c (create_wide_cs): New static function. (wide_display_char_p): New function. * regex.h (wide_display_char_p): Declared. * stream.c (put_string, put_char): Use wide_display_char_p to determine whether an extra column need be counted. Also bugfix: iswprint evidently cannot be relied to work over the entire Unicode range, at least not in the C locale. Glibc's version and is reporting valid Japanese characters as unprintable on Ubuntu. As a hack we instead check for control characters and invert the result: control chars are unprintable. * tests/009/json.expected: Updated.
* Handle abc: token syntax.Kaz Kylheku2015-08-101-2/+2
| | | | | | * parser.l (BTREG, NTREG): Allow an empty string symbol name with a nonempty package name. Without this, abc: parses as abc :.
* Diagnose bad consing dot syntax like (a . b . c).Kaz Kylheku2015-08-104-9/+27
| | | | | | | | | | | | | | | | | | | | | * parser.y (r_exprs): Use unique object in the terminating cons to indicate the empty spot where the dotted cdr item will go. Check for misplaced consing dot. (misplaced_consing_dot_check): New static function. Checks for the terminator atom spot being taken already. Thus, the spot may be taken only by the very last reduction, such that the next reduction is r_exprs -> n_exprs where the terminating atom is processed. * parser.c (unique_s): New global variable. (parse_init): Initialize unique_s. * parser.h (unique_s): Declared. * share/txr/stdlib/place.tl (sys:placelet-1): We have a misplaced consing dot here! It was working correctly by "terminating atom propagation" behavior, which allowed (a . b c d) to produce (a c d . b). If a single terminating atom occurred in the middle of a list, it was promoted to the end.
* Remove never-used member of symbol structure.Kaz Kylheku2015-08-093-3/+0
| | | | | | | | * lib.h (struct sym): Remove value member. * lib.c (make_sym): Don't initialize removed member. * gc.c (mark_obj): Do not mark removed member.
* Discontinuing ChangeLog.Kaz Kylheku2015-08-081-0/+8
| | | | | | * ChangeLog: renamed to ChangeLog-2009-2015, and discontinued. Changes will be tracked only in the git commit messages from now on and not duplicated into the ChangeLog.
* Version 111.txr-111Kaz Kylheku2015-08-087-210/+311
|
* C++ upkeep: conversions, clashes, warnings.Kaz Kylheku2015-08-074-5/+16
| | | | | | | | | * glob.c (glob_wrap): Fix signed/unsigned comparison. * stream.c (make_null_stream): Fix convert beign used where coerce is needed. * sysif.c (dup_wrap): Fix use of C++ new keyword.
* C++ static forward issue.Kaz Kylheku2015-08-073-3/+21
| | | | | | | | * lib.h (static_forward, static_def): New macros for dealing with C++ static forward declaration problem. * syslog.c (syslog_strm_ops): Use static forward macros.
* C++ upkeep: resolve multiple definitions of fun_k.Kaz Kylheku2015-08-076-7/+23
| | | | | | | | | | | | | | | * eval.c (fun_k): Global definition removed. (eval_init): Do not initialize fun_k here. * filter.c (fun_k): Definition removed. (filter_init): Do not initialize fun_k. * filter.h (fun_k): Declaration removed. * lib.c (fun_k): Defined in this file now. (obj_init): Initialize fun_k here. * lib.h (fun_k): Declare here.
* * Makefile (SRCS): When top_srcdir is blank, elide the entireKaz Kylheku2015-08-072-1/+8
| | | | | --work-tree argument to git. On Cygwin, git throws a strange error message when --work-tree is given a blank argument.
generated by cgit v1.2.3 (git 2.25.1) at 2025-03-16 14:15:49 +0000