jamgram.h #1 

#define _BANG 257
#define _BANG_EQUALS 258
#define _AMPERAMPER 259
#define _LPAREN 260
#define _RPAREN 261
#define _PLUS_EQUALS 262
#define _COLON 263
#define _SEMIC 264
#define _LANGLE 265
#define _LANGLE_EQUALS 266
#define _EQUALS 267
#define _RANGLE 268
#define _RANGLE_EQUALS 269
#define _QUESTION_EQUALS 270
#define ACTIONS 271
#define BIND 272
#define CASE 273
#define DEFAULT 274
#define ELSE 275
#define EXISTING 276
#define FOR 277
#define IF 278
#define IGNORE 279
#define IN 280
#define INCLUDE 281
#define LOCAL 282
#define ON 283
#define PIECEMEAL 284
#define QUIETLY 285
#define RULE 286
#define SWITCH 287
#define TOGETHER 288
#define UPDATED 289
#define _LBRACE 290
#define _BARBAR 291
#define _RBRACE 292
#define ARG 293
#define STRING 294
# Change User Description Committed
#4 1270 Matt Armstrong In prep for publishing all my patches to jam.
#3 1088 Matt Armstrong Merge in changes from upstream.
#2 293 Matt Armstrong Fix a crash bug in jam by adding reference counting to the
PARSE struct.  The bug occurs when a rule re-defines
itself.  This can only happen by using the include statement,
as follows.

JamfileA:
----------------------------------------------------
rule DoInclude {
    ECHO JamfileA DoInclude $(<) start ;
    include $(<) ;
    ECHO JamfileA DoInclude $(<) end ;
}
DoInclude JambileB ;
----------------------------------------------------

JamfileB:
----------------------------------------------------
rule DoInclude {
    ECHO JamfileB DoInclude ;
}
DoInclude JambileC ;
----------------------------------------------------

With these two jamfiles you can reproduce the crash bug
by running "jam -fJamfileA".  If that doesn't work,
change parse_free() such that it memsets PARSE struct being
freed to something bogus like 0x03 and try again (or use
Purify and watch the free memory read warnings).

The problem is that compile_setcomp() just frees the PARSE
without regard for the possibility that it might be executing.
In the case above, JamfileA's DoInclude is re-defined by
JamfileB while it is still being executed.

The simple fix is to just have compile_setcomp() not free the
PARSE.  This would introduce a memory leak for every re-defined
rule.

My fix introduces reference counting of the PARSE struct.  A
PARSE is allocated with a reference count of 1 and when it
goes to 0 it is freed.  Every place that was a parse_free() is
now a parse_decref().  In addition, compile_rules() increments
the reference count of its PARSE around the calls to its left
and right functions.

Perhaps there a better fix!
#1 289 Matt Armstrong Initial branch.
//guest/perforce_software/jam/src/jamgram.h
#1 2 laura Add Jam/MR 2.2 source