2 =======================================================
3 Known Problems In PCCTS - Last revised 14 November 1998
4 =======================================================
6 #14. Parsing bug in dlg
8 THM: I have been unable to reproduce this problem.
10 Reported by Rick Howard Mijenix Corporation (rickh@mijenix.com).
12 The regular expression parser (in rexpr.c) fails while
13 trying to parse the following regular expression:
15 {[a-zA-Z]:}(\\\\[a-zA-Z0-9]*)+
17 See my comment in the following excerpt from rexpr.c:
20 * <regExpr> ::= <andExpr> ( '|' {<andExpr>} )*
22 * Return -1 if syntax error
23 * Return 0 if none found
24 * Return 1 if a regExrp was found
32 if ( andExpr(&g1) == -1 )
37 while ( token == '|' )
42 if ( a == -1 ) return -1; /* syntax error below */
43 else if ( !a ) return 1; /* empty alternative */
44 g1 = BuildNFA_AorB(g1, g2);
47 if ( token!='\0' ) return -1;
49 ***** It appears to fail here becuause token is 125 - the closing '}'
50 ***** If I change it to:
51 ***** if ( token!='\0' && token!='}' && token!= ')' ) return -1;
53 ***** It succeeds, but I'm not sure this is the corrrect approach.
59 #13. dlg reports an invalid range for: [\0x00-\0xff]
61 Diagnosed by Piotr Eljasiak (eljasiak@no-spam.zt.gdansk.tpsa.pl):
65 #12. Strings containing comment actions
67 Sequences that looked like C style comments appearing in string
68 literals are improperly parsed by antlr/dlg.
70 << fprintf(out," /* obsolete */ ");
74 << fprintf(out," \/\* obsolete \*\/ ");
76 Reported by K.J. Cummings (cummings@peritus.com).
78 #11. User hook for deallocation of variables on guess fail
80 The mechanism outlined in Item #108 works only for
81 heap allocated variables.
83 #10. Label re-initialization in ( X {y:Y} )*
85 If a label assignment is optional and appears in a
86 (...)* or (...)+ block it will not be reset to NULL
87 when it is skipped by a subsequent iteration.
97 The first time through the block Y will be matched and
98 y will be set to point to the token. On the second
99 iteration of the (...)* block there is no match for Y.
100 But y will not be reset to NULL, as the user might
101 expect, it will contain a reference to the Y that was
102 matched in the first iteration.
104 The work-around is to manually reset y:
106 ( X << y = NULL; >> { y:Y } )* Z
110 ( X ( y:Y | << y = NULL; >> /* epsilon */ ) )* Z
112 Reported by Jeff Vincent (JVincent@novell.com).
114 #9. PCCTAST.h PCCTSAST::setType() is a noop
116 #8. #tokdefs with ~Token and .
118 THM: I have been unable to reproduce this problem.
120 When antlr uses #tokdefs to define tokens the fields of
121 #errclass and #tokclass do not get properly defined.
122 When it subsequently attempts to take the complement of
123 the set of tokens (using ~Token or .) it can refer to
124 tokens which don't have names, generating a fatal error.
126 #7. DLG crashes on some invalid inputs
128 THM: In MR20 have fixed the most common cases.
130 The following token defintion will cause DLG to crash.
134 Reported by Mengue Olivier (dolmen@bigfoot.com).
136 #6. On MS systems \n\r is treated as two new lines
140 #5. Token expressions in #tokclass
142 #errclass does not support TOK1..TOK2 or ~TOK syntax.
143 #tokclass does not support ~TOKEN syntax
145 A workaround for #errclass TOK1..TOK2 is to use a
148 Reported by Dave Watola (dwatola@amtsun.jpl.nasa.gov)
150 #4. A #tokdef must appear "early" in the grammar file.
152 The "early" section of the grammar file is the only
153 place where the following directives may appear:
160 Any other kind of statement signifiies the end of the
163 #3. Use of PURIFY macro for C++ mode
165 Item #93 of the CHANGES_FROM_1.33 describes the use of
166 the PURIFY macro to zero arguments to be passed by
169 #define PURIFY(r, s) memset((char *) &(r), '\0', (s));
171 This may not be the right thing to do for C++ objects that
172 have constructors. Reported by Bonny Rais (bonny@werple.net.au).
174 For those cases one should #define PURIFY to be an empty macro
175 in the #header or #first actions.
177 #2. Fixed in 1.33MR10 - See CHANGES_FROM_1.33 Item #80.
179 #1. The quality of support for systems with 8.3 file names leaves
180 much to be desired. Since the kit is distributed using the
181 long file names and the make file uses long file names it requires
182 some effort to generate. This will probably not be changed due
183 to the large number of systems already written using the long