From e73034b0d2191f00a8b00f38383843ea8858c576 Mon Sep 17 00:00:00 2001
From: mdw <mdw>
Date: Thu, 19 Aug 1999 18:32:48 +0000
Subject: [PATCH] Improve lexical analysis.  In particular, `chmod' patterns
 don't have to be quoted any more.

---
 conf.c  | 34 ++++++++++++++++++++++++++++------
 conf.h  | 19 ++++++++++++++++++-
 fattr.c | 35 +++++++++++++++++++++--------------
 fw.1    | 15 ++++++++-------
 4 files changed, 75 insertions(+), 28 deletions(-)

diff --git a/conf.c b/conf.c
index 020df7a..73b611f 100644
--- a/conf.c
+++ b/conf.c
@@ -1,6 +1,6 @@
 /* -*-c-*-
  *
- * $Id: conf.c,v 1.2 1999/07/26 23:28:39 mdw Exp $
+ * $Id: conf.c,v 1.3 1999/08/19 18:32:48 mdw Exp $
  *
  * Configuration parsing
  *
@@ -29,6 +29,10 @@
 /*----- Revision history --------------------------------------------------* 
  *
  * $Log: conf.c,v $
+ * Revision 1.3  1999/08/19 18:32:48  mdw
+ * Improve lexical analysis.  In particular, `chmod' patterns don't have to
+ * be quoted any more.
+ *
  * Revision 1.2  1999/07/26 23:28:39  mdw
  * Major reconstruction work for new design.
  *
@@ -66,8 +70,24 @@
 static source_ops *sources[] = { &xsource_ops, &fsource_ops, &ssource_ops };
 static target_ops *targets[] = { &xtarget_ops, &ftarget_ops, &starget_ops };
 
+static const char *notword = 0;
+static const char *notdelim = 0;
+
 /*----- Main code ---------------------------------------------------------*/
 
+/* --- @undelim@ --- *
+ *
+ * Arguments:	@const char *d, dd@ = pointer to characters to escape
+ *
+ * Returns:	---
+ *
+ * Use:		Modifies the tokenizer.  Characters in the first list will
+ *		always be considered to begin a word.  Characters in the
+ *		second list will always be allowed to continue a word.
+ */
+
+void undelim(const char *d, const char *dd) { notword = d; notdelim = dd; }
+
 /* --- @token@ --- *
  *
  * Arguments:	@scanner *sc@ = pointer to scanner definition
@@ -122,10 +142,12 @@ int token(scanner *sc)
       /* --- Various self-delimiting characters --- */
 
       case SELFDELIM:
-	dstr_putc(&sc->d, ch);
-	dstr_putz(&sc->d);
-	sc->t = ch;
-	goto done;
+	if (!notword || strchr(notword, ch) == 0) {
+	  dstr_putc(&sc->d, ch);
+	  dstr_putz(&sc->d);
+	  sc->t = ch;
+	  goto done;
+	}
 
       /* --- Bare words --- *
        *
@@ -150,7 +172,7 @@ int token(scanner *sc)
 	      q = !q;
 	      break;
 	    case SELFDELIM:
-	      if (q)
+	      if (q || (notdelim && strchr(notdelim, ch)))
 		goto insert;
 	      goto word;
 	    default:
diff --git a/conf.h b/conf.h
index 6183048..e0eddf8 100644
--- a/conf.h
+++ b/conf.h
@@ -1,6 +1,6 @@
 /* -*-c-*-
  *
- * $Id: conf.h,v 1.3 1999/07/27 18:30:14 mdw Exp $
+ * $Id: conf.h,v 1.4 1999/08/19 18:32:48 mdw Exp $
  *
  * Configuration parsing
  *
@@ -29,6 +29,10 @@
 /*----- Revision history --------------------------------------------------* 
  *
  * $Log: conf.h,v $
+ * Revision 1.4  1999/08/19 18:32:48  mdw
+ * Improve lexical analysis.  In particular, `chmod' patterns don't have to
+ * be quoted any more.
+ *
  * Revision 1.3  1999/07/27 18:30:14  mdw
  * Improve documentation and layout for @CONF_BEGIN@ and friends.  Remove
  * irritating warning about unused label by introducing a spurious `goto'.
@@ -63,6 +67,19 @@
 
 /*----- Functions provided ------------------------------------------------*/
 
+/* --- @undelim@ --- *
+ *
+ * Arguments:	@const char *d, dd@ = pointer to characters to escape
+ *
+ * Returns:	---
+ *
+ * Use:		Modifies the tokenizer.  Characters in the first list will
+ *		always be considered to begin a word.  Characters in the
+ *		second list will always be allowed to continue a word.
+ */
+
+extern void undelim(const char */*d*/, const char */*dd*/);
+
 /* --- @token@ --- *
  *
  * Arguments:	@scanner *sc@ = pointer to scanner definition
diff --git a/fattr.c b/fattr.c
index 0406a11..75491a6 100644
--- a/fattr.c
+++ b/fattr.c
@@ -1,6 +1,6 @@
 /* -*-c-*-
  *
- * $Id: fattr.c,v 1.1 1999/07/26 23:33:56 mdw Exp $
+ * $Id: fattr.c,v 1.2 1999/08/19 18:32:48 mdw Exp $
  *
  * Handling of file attributes
  *
@@ -29,6 +29,10 @@
 /*----- Revision history --------------------------------------------------* 
  *
  * $Log: fattr.c,v $
+ * Revision 1.2  1999/08/19 18:32:48  mdw
+ * Improve lexical analysis.  In particular, `chmod' patterns don't have to
+ * be quoted any more.
+ *
  * Revision 1.1  1999/07/26 23:33:56  mdw
  * Support code for new design.
  *
@@ -100,12 +104,14 @@ int fattr_option(scanner *sc, fattr *f)
 
     /* --- Gobble an optional `=' sign --- */
 
+    undelim(0, ",=+-");
     token(sc);
     if (sc->t == '=')
       token(sc);
 
     if (sc->t != CTOK_WORD)
       error(sc, "parse error, expected file mode");
+    undelim(0, 0);
 
     /* --- If it looks digitlike, read as octal --- */
 
@@ -116,7 +122,8 @@ int fattr_option(scanner *sc, fattr *f)
 
     else {
       const char *p;
-      unsigned mask = 07777;
+      unsigned mask = 04700;
+      unsigned state = 0;
       unsigned or = 1;
 
       /* --- Set the default from the umask --- */
@@ -137,21 +144,21 @@ int fattr_option(scanner *sc, fattr *f)
 	switch (*p) {
 	  case ',': break;
 
-	  case 'a': mask = 07777; break;
-	  case 'u': mask = 04700; break;
-	  case 'g': mask = 02070; break;
-	  case 'o': mask = 01007; break;
+	  case 'a': mask = (mask & state) | 07777; state = 07777; break;
+	  case 'u': mask = (mask & state) | 04700; state = 07777; break;
+	  case 'g': mask = (mask & state) | 02070; state = 07777; break;
+	  case 'o': mask = (mask & state) | 01007; state = 07777; break;
 
-	  case '=': mode &= ~mask; break;
-	  case '-': or = 0; break;
-	  case '+': or = 1; break;
+	  case '=': mode &= ~mask; /* Drop through */
+	  case '+': state = 0; or = 1; break;
+	  case '-': state = 0; or = 0; break;
 
 #define APPLY(m) if (or) mode |= ((m) & mask); else mode &= ~((m) & mask);
-	  case 'r': APPLY(00444); break;
-	  case 'w': APPLY(00222); break;
-	  case 'x': APPLY(00111); break;
-	  case 's': APPLY(06000); break;
-	  case 't': APPLY(01000); break;
+	  case 'r': state = 0; APPLY(00444); break;
+	  case 'w': state = 0; APPLY(00222); break;
+	  case 'x': state = 0; APPLY(00111); break;
+	  case 's': state = 0; APPLY(06000); break;
+	  case 't': state = 0; APPLY(01000); break;
 #undef APPLY
 
 	  default: error(sc, "unknown mode character `%c'", *p);
diff --git a/fw.1 b/fw.1
index c65c6c0..6cd0c64 100644
--- a/fw.1
+++ b/fw.1
@@ -1,6 +1,6 @@
 .\" -*-nroff-*-
 .\"
-.\" $Id: fw.1,v 1.3 1999/07/30 06:49:00 mdw Exp $
+.\" $Id: fw.1,v 1.4 1999/08/19 18:32:48 mdw Exp $
 .\"
 .\" Manual page for fw
 .\"
@@ -28,6 +28,10 @@
 .\" ---- Revision history ---------------------------------------------------
 .\" 
 .\" $Log: fw.1,v $
+.\" Revision 1.4  1999/08/19 18:32:48  mdw
+.\" Improve lexical analysis.  In particular, `chmod' patterns don't have to
+.\" be quoted any more.
+.\"
 .\" Revision 1.3  1999/07/30 06:49:00  mdw
 .\" Minor tidying and typo correction.
 .\"
@@ -482,7 +486,7 @@ sources and targets is like this:
 .br
 .I null-spec
 ::=
-.RB [[ : ] null [ : ]]
+.RB [ : ] null [ : ]
 .PP
 The
 .I file
@@ -603,14 +607,11 @@ argument may be either an octal number or a
 string which acts on the default permissions established by the
 prevailing
 .BR umask (2)
-setting.  Note that
-.BR chmod -style
-strings may contain
+setting.  The characters
 .RB ` = '
 and
 .RB ` , '
-characters that will need to be escaped or quoted.
-.RE
+do not have to be quoted within the mode string.
 .PP
 .IB prefix .fattr.owner
 .RB [ = ]
-- 
2.11.0