[9fans] regexp(7): too many character classes; limit@ -- tried to match sha1

[dexen deVries]

a plumber rule to match an sha1 hash by catenating 40 of '[0-9a-f]' character
classes. got ``too many character classes; limit@'' error.


so far i've found
#define NCLASS (sizeof(RePrOg.class)/sizeof(Reclass))
but no definition of RePrOg, only
extern Reprog RePrOg;


how do i work around it?

[erik quanstrom]

Post by dexen deVries
a plumber rule to match an sha1 hash by catenating 40 of '[0-9a-f]' character
so far i've found
#define NCLASS (sizeof(RePrOg.class)/sizeof(Reclass))
but no definition of RePrOg, only
extern Reprog RePrOg;

yes, that is quite an obtuse definition. i had a similar problem
in plan 9 and changed things so the definitions were explicit
in regexp.h. at the same time, i noticed that the test for adjacent
ranges was incorrect, but that doesn't appear to be your problem.

i put the full source here:

/n/sources/contrib/quanstro/regexp.tar

not every difference is a good one. :-) it looks like i need to
do a bit of cleanup.

- erik

---

/n/sources/plan9/sys/include/regexp.h:1,6 - /sys/include/regexp.h:1,11
#pragma src "/sys/src/libregexp"
#pragma lib "libregexp.a"

+ enum {
+ NSPANS = 128, /* max rune ranges per character class */
+ NCLASS = 16, /* max character classes per program */
+ };
+
typedef struct Resub Resub;
typedef struct Reclass Reclass;
typedef struct Reinst Reinst;
/n/sources/plan9/sys/include/regexp.h:27,33 - /sys/include/regexp.h:32,38
*/
struct Reclass{
Rune *end;
- Rune spans[64];
+ Rune spans[NSPANS*2];
};

/*
/n/sources/plan9/sys/include/regexp.h:52,58 - /sys/include/regexp.h:57,63
*/
struct Reprog{
Reinst *startinst; /* start pc */
- Reclass class[16]; /* .data */
+ Reclass class[NCLASS]; /* .data */
Reinst firstinst[5]; /* .text */
};

/n/sources/plan9/sys/src/libregexp/regcomp.c:389,399 - regcomp.c:383,408
static int
bldcclass(void)
{
int type;
- Rune r[NCCRUNE];
+ Rune r[NSPANS*2];
Rune *p, *ep, *np;
Rune rune;
int quoted;