git: 9front

Info  •  Files  •  Log  •  Branches

libregexp: fix handling of empty regprog //

Instead of making empty regprog equal to ^
make the empty regprog match every position
in the string with a zero-length match.

This fixes using // as separator in awk's
gsub operation.

--- a/sys/man/6/regexp

+++ b/sys/man/6/regexp

@@ -91,8 +91,6 @@

 matches the beginning of a line;

 .L $

 matches the end of the line.

-An empty regular expression behaves equivalently to

-.LR ^ .

 .PP

 The 

 .B REP

--- a/sys/src/libregexp/regcomp.c

+++ b/sys/src/libregexp/regcomp.c

@@ -163,13 +163,9 @@

 	Renode *parsetr;

 	int regstrlen, maxthr;

-	regstrlen = utflen(regstr);

-	if(regstrlen == 0){

-		regstr = "^";

-		regstrlen = 1;

-	}

 	initplex(&plex, regstr, lit);

-	plex.nodes = calloc(sizeof(*plex.nodes), regstrlen*2);

+	regstrlen = utflen(regstr);

+	plex.nodes = calloc(sizeof(*plex.nodes), regstrlen*2+1);

 	if(plex.nodes == nil)

 		return nil;

 	plex.next = plex.nodes;

@@ -180,7 +176,7 @@

 	}

 	maxthr = regstrlen + 1;

-	parsetr = node(&plex, TSUB, e0(&plex), nil);

+	parsetr = node(&plex, TSUB, regstrlen?e0(&plex):nil, nil);

 //	prtree(parsetr, 0, 1);

 	reprog = malloc(sizeof(Reprog) +

-- 

⑨