TODO

   1 -*- outline -*-
   2
   3 * Unit rules
   4 Maybe we could expand unit rules, i.e., transform
   5
   6         exp: arith | bool;
   7         arith: exp '+' exp;
   8         bool: exp '&' exp;
   9
  10 into
  11
  12         exp: exp '+' exp | exp '&' exp;
  13
  14 when there are no actions.  This can significantly speed up some
  15 grammars.
  16
  17 * Huge Grammars
  18 Currently, not only is Bison unable to handle huge grammars because of
  19 internal limitations, but the test `big triangle' also demonstrates
  20 that it can produce SEGVing executables!  Push the limit beyond 124,
  21 and have a core dump.  Be my guest: fix this!
  22
  23 * read_pipe.c
  24 This is not portable to DOS for instance.  Implement a more portable
  25 scheme.  Sources of inspiration include GNU diff, and Free Recode.
  26
  27 * NEWS
  28 Sort from 1.31 NEWS.
  29
  30 * Prologue
  31 The %union is declared after the user C declarations. It can be
  32 a problem if YYSTYPE is declared after the user part.   []
  33
  34 Actually, the real problem seems that the %union ought to be output
  35 where it was defined.  For instance, in gettext/intl/plural.y, we
  36 have:
  37
  38         %{
  39         ...
  40         #include "gettextP.h"
  41         ...
  42         %}
  43
  44         %union {
  45           unsigned long int num;
  46           enum operator op;
  47           struct expression *exp;
  48         }
  49
  50         %{
  51         ...
  52         static int yylex PARAMS ((YYSTYPE *lval, const char **pexp));
  53         ...
  54         %}
  55
  56 Where the first part defines struct expression, the second uses it to
  57 define YYSTYPE, and the last uses YYSTYPE.  Only this order is valid.
  58
  59 * --graph
  60 Show reductions.        []
  61
  62 * Broken options ?
  63 ** %no-lines            [ok]
  64 ** %no-parser           []
  65 ** %pure-parser         []
  66 ** %semantic-parser     []
  67 ** %token-table         []
  68 ** Options which could use parse_dquoted_param ().
  69 Maybe transfered in lex.c.
  70 *** %skeleton           [ok]
  71 *** %output             []
  72 *** %file-prefix        []
  73 *** %name-prefix        []
  74
  75 ** Skeleton strategy.   []
  76 Must we keep %no-parser?
  77              %token-table?
  78 *** New skeletons.      []
  79
  80 * src/print_graph.c
  81 Find the best graph parameters. []
  82
  83 * doc/bison.texinfo
  84 ** Update
  85 informations about ERROR_VERBOSE.       []
  86 ** Add explainations about
  87 skeleton muscles.       []
  88 %skeleton.              []
  89
  90 * testsuite
  91 ** tests/pure-parser.at []
  92 New tests.
  93
  94 * Debugging parsers
  95
  96 From Greg McGary:
  97
  98 akim demaille <akim.demaille@epita.fr> writes:
  99
 100 > With great pleasure!  Nonetheless, things which are debatable
 101 > (or not, but just `big') should be discuss in `public': something
 102 > like help- or bug-bison@gnu.org is just fine.  Jesse and I are there,
 103 > but there is also Jim and some other people.
 104
 105 I have no idea whether it qualifies as big or controversial, so I'll
 106 just summarize for you.  I proposed this change years ago and was
 107 surprised that it was met with utter indifference!
 108
 109 This debug feature is for the programs/grammars one develops with
 110 bison, not for debugging bison itself.  I find that the YYDEBUG
 111 output comes in a very inconvenient format for my purposes.
 112 When debugging gcc, for instance, what I want is to see a trace of
 113 the sequence of reductions and the line#s for the semantic actions
 114 so I can follow what's happening.  Single-step in gdb doesn't cut it
 115 because to move from one semantic action to the next takes you through
 116 lots of internal machinery of the parser, which is uninteresting.
 117
 118 The change I made was to the format of the debug output, so that it
 119 comes out in the format of C error messages, digestible by emacs
 120 compile mode, like so:
 121
 122 grammar.y:1234: foo: bar(0x123456) baz(0x345678)
 123
 124 where "foo: bar baz" is the reduction rule, whose semantic action
 125 appears on line 1234 of the bison grammar file grammar.y.  The hex
 126 numbers on the rhs tokens are the parse-stack values associated with
 127 those tokens.  Of course, yytype might be something totally
 128 incompatible with that representation, but for the most part, yytype
 129 values are single words (scalars or pointers).  In the case of gcc,
 130 they're most often pointers to tree nodes.  Come to think of it, the
 131 right thing to do is to make the printing of stack values be
 132 user-definable.  It would also be useful to include the filename &
 133 line# of the file being parsed, but the main filename & line# should
 134 continue to be that of grammar.y
 135
 136 Anyway, this feature has saved my life on numerous occasions.  The way
 137 I customarily use it is to first run bison with the traces on, isolate
 138 the sequence of reductions that interests me, put those traces in a
 139 buffer and force it into compile-mode, then visit each of those lines
 140 in the grammar and set breakpoints with C-x SPACE.  Then, I can run
 141 again under the control of gdb and stop at each semantic action.
 142 With the hex addresses of tree nodes, I can inspect the values
 143 associated with any rhs token.
 144
 145 You like?
 146
 147 * input synclines
 148 Some users create their foo.y files, and equip them with #line.  Bison
 149 should recognize these, and preserve them.