Stripping of Xo/Xc macros in main.c (ifdef STRIP_XO).

4 files changed, 136 insertions, 65 deletions

diff --git a/Makefile b/Makefile
index b7d83b37..f186e817 100644
--- a/Makefile
+++ b/Makefile
@@ -13,10 +13,14 @@ VERSION	   = 1.6.8
 VDATE	   = 21 March 2009
 
 VFLAGS     = -DVERSION=\"$(VERSION)\"
-CFLAGS    += -W -Wall -Wstrict-prototypes -Wno-unused-parameter -g 
+CFLAGS    += -W -Wall -Wstrict-prototypes -Wno-unused-parameter -g
 LINTFLAGS += $(VFLAGS)
 CFLAGS    += $(VFLAGS)
 
+# If you want to strip `Xo/Xc' macro pairs, enable this.  Really, only
+# OpenBSD should be doing this while it kicks its cruft.
+CFLAGS	  += -DSTRIP_XO
+
 LIBLNS	   = macro.ln mdoc.ln hash.ln strings.ln xstd.ln argv.ln \
 	     validate.ln action.ln lib.ln att.ln arch.ln vol.ln \
 	     msec.ln st.ln
diff --git a/macro.c b/macro.c
index 9549004d..6b3e4fe3 100644
--- a/macro.c
+++ b/macro.c
@@ -1,4 +1,4 @@
-/* $Id: macro.c,v 1.76 2009/03/21 09:42:07 kristaps Exp $ */
+/* $Id: macro.c,v 1.77 2009/03/22 19:01:11 kristaps Exp $ */
 /*
  * Copyright (c) 2008, 2009 Kristaps Dzonsons <kristaps@openbsd.org>
  *
@@ -33,6 +33,7 @@
 /* FIXME: .Fl, .Ar, .Cd handling of `|'. */
 
 enum	mwarn {
+	WIMPBRK,
 	WMACPARM,
 	WOBS
 };
@@ -236,6 +237,9 @@ pwarn(struct mdoc *mdoc, int line, int pos, enum mwarn type)
 
 	p = NULL;
 	switch (type) {
+	case (WIMPBRK):
+		p = "crufty end-of-line scope violation";
+		break;
 	case (WMACPARM):
 		p = "macro-like parameter";
 		break;
@@ -1058,6 +1062,9 @@ blk_part_imp(MACRO_PROT_ARGS)
 		if (body == n)
 			break;
 
+	if (NULL == n && ! pwarn(mdoc, body->line, body->pos, WIMPBRK))
+			return(0);
+
 	if (n && ! rew_last(mdoc, body))
 		return(0);
 
diff --git a/main.c b/main.c
index 773700e9..889b3387 100644
--- a/main.c
+++ b/main.c
@@ -1,4 +1,4 @@
-/* $Id: main.c,v 1.7 2009/03/20 21:58:38 kristaps Exp $ */
+/* $Id: main.c,v 1.8 2009/03/22 19:01:11 kristaps Exp $ */
 /*
  * Copyright (c) 2008, 2009 Kristaps Dzonsons <kristaps@openbsd.org>
  *
@@ -264,6 +264,9 @@ fdesc(struct buf *blk, struct buf *ln,
 	ssize_t		 ssz;
 	struct stat	 st;
 	int		 j, i, pos, lnn;
+#ifdef	STRIP_XO
+	int		 macro, xo, xeoln;
+#endif
 
 	/*
 	 * Two buffers: ln and buf.  buf is the input buffer, optimised
@@ -288,6 +291,9 @@ fdesc(struct buf *blk, struct buf *ln,
 	/*
 	 * Fill buf with file blocksize and parse newlines into ln.
 	 */
+#ifdef	STRIP_XO
+	macro = xo = xeoln = 0;
+#endif
 
 	for (lnn = 1, pos = 0; ; ) {
 		if (-1 == (ssz = read(fd, blk->buf, sz))) {
@@ -305,6 +311,66 @@ fdesc(struct buf *blk, struct buf *ln,
 			}
 
 			if ('\n' != blk->buf[i]) {
+				/*
+				 * Ugly of uglies.  Here we handle the
+				 * dreaded `Xo/Xc' scoping.  Cover the
+				 * eyes of any nearby children.  This
+				 * makes `Xo/Xc' enclosures look like
+				 * one huge line.
+				 */
+#ifdef	STRIP_XO
+				/*
+				 * First, note whether we're in a macro
+				 * line.
+				 */
+				if (0 == pos && '.' == blk->buf[i])
+					macro = 1;
+
+				/*
+				 * If we're in an `Xo' context and just
+				 * nixed a newline, remove the control
+				 * character for new macro lines:
+				 * they're going to show up as all part
+				 * of the same line.
+				 */
+				if (xo && xeoln && '.' == blk->buf[i]) {
+					xeoln = 0;
+					continue;
+				}
+				xeoln = 0;
+
+				/*
+				 * If we've parsed `Xo', enter an xo
+				 * context.  `Xo' must be in a parsable
+				 * state.  This is the ugly part.  IT IS
+				 * NOT SMART ENOUGH TO HANDLE ESCAPED
+				 * WHITESPACE.
+				 */
+				if (macro && pos && 'o' == blk->buf[i]) {
+					if (xo && 'X' == ln->buf[pos - 1])  {
+						if (' ' == ln->buf[pos - 2])
+							xo++;
+					} else if ('X' == ln->buf[pos - 1]) {
+						if (2 == pos && '.' == ln->buf[pos - 2])
+							xo++;
+						else if (' ' == ln->buf[pos - 2])
+							xo++;
+					}
+				}
+
+				/*
+				 * If we're parsed `Xc', leave an xo
+				 * context if one's already pending.
+				 * `Xc' must be in a parsable state.
+				 * THIS IS NOT SMART ENOUGH TO HANDLE
+				 * ESCAPED WHITESPACE.
+				 */
+				if (macro && pos && 'c' == blk->buf[i])
+					if (xo && 'X' == ln->buf[pos - 1])
+						if (' ' == ln->buf[pos - 2])
+							xo--;
+#endif	/* STRIP_XO */
+
 				ln->buf[pos++] = blk->buf[i];
 				continue;
 			}
@@ -323,11 +389,25 @@ fdesc(struct buf *blk, struct buf *ln,
 				}
 			}
 
+#ifdef	STRIP_XO
+			/*
+			 * If we're in an xo context, put a space in
+			 * place of the newline and continue parsing.
+			 * Mark that we just did a newline.
+			 */
+			if (xo) {
+				xeoln = 1;
+				ln->buf[pos++] = ' ';
+				lnn++;
+				continue;
+			}
+#endif	/* STRIP_XO */
+
 			ln->buf[pos] = 0;
 			if ( ! mdoc_parseln(mdoc, lnn, ln->buf))
 				return(0);
 			lnn++;
-			pos = 0;
+			macro = pos = 0;
 		}
 	}
 
diff --git a/manuals.7 b/manuals.7
index 01301213..5ce9b645 100644
--- a/manuals.7
+++ b/manuals.7
@@ -25,22 +25,20 @@ documentation
 If you add something to your operating system, whether it's a new file
 format or directory structure or device driver, it needs documentation.
 .\" SECTION
-.Sh CLASSIFICATION
-Classify your system component.  In
-.Ux ,
-each component has a manual section , which categorises the component's
-function.  The section of a manual is usually listed in parenthesis next
-to the component name, such as
-.Xr ps 1 ,
-section 1.  You can query a manual explicitly by its section:
-.Bd -literal -offset XXXX
-% man \-s 1 ps
-% apropos ps
-.Ed
-.Pp
-The following table lists classifications and the applicable manual
-sections:
+.Sh COMPOSITION
+Prepare your composition environment by copying over the manual template
+from 
+.Pa /usr/share/misc/mdoc.template .
+.Em \&Do not
+start afresh or by copying another manual unless you know exactly what
+you're doing!
+.\" SUBSECTION
+.Ss Section Numbering
+Find an appropriate section for your manual.  There may exist multiple
+manual names per section, so be specific.  A table of all available
+manual sections follows:
 .Pp
+.\" LIST
 .Bl -tag -width "XXXXXXXXXXXX" -offset indent -compact
 .It Em Section
 .Em Description
@@ -62,52 +60,30 @@ administrator utilities
 in-kernel routines
 .El
 .Pp
-Some examples in regular name/section form:
-.Pp
-.\" LIST
-.Bl -tag -width "File-formatX" -offset indent -compact
-.It Em Manual
-.Em Description
-.It Xr dc 4
-DEC/Intel 10/100 Ethernet device
-.It Xr usermod 8
-modify user login information
-.It Xr cc 1
-the C compiler
-.It Xr fortune 6
-print a random adage
-.It Xr open 2
-open or create a file for reading or writing
-.It Xr isspace 3
-whitespace character test
-.It Xr Pod::Man 3p
-convert POD data to formatted roff
-.It Xr tsleep 9
-process context sleep
-.It Xr passwd 5
-format of the password file
-.It Xr symlink 7
-symbolic link handling
-.El
-.\" SECTION
-.Sh COMPOSITION
-Prepare your composition environment by copying over the manual template
-from 
-.Pa /usr/share/misc/mdoc.template .
-.Em \&Do not
-start afresh or by copying another manual unless you know exactly what
-you're doing!
+If your manual falls into multiple categories, choose the most
+widely-used or, better, re-consider the topic of your manual to be more
+specific.  You can list all manuals per section by invoking
+.Xr apropos 1 ,
+then provide the
+.Fl s
+flag to
+.Xr man 1
+to see the specific section manual (section 1, in this example):
+.\" DISPLAY
+.Bd -literal -offset indent
+% apropos myname
+myname (1) - some description here
+% man \-s 1 myname
+.Ed
 .\" SUBSECTION
 .Ss Naming
-Your component will need a name by which to query its contents via
-.Xr man 1 .
-Keep it simple.  You may want to look for other manuals by that same
-name before committing:
+Name your component.  Be terse, erring on the side of clarity.  You may
+want to look for other manuals by that same name before committing:
 .Pp
 .Dl % apropos myname
 .Pp
-Conventionally, manual files are named 
-.Pa myname.section ,
+Manual files are named 
+.Pa myname.mysection ,
 such as
 .Pa manuals.7
 for this document.
@@ -129,13 +105,13 @@ packages of
 .Xr roff 7 ;
 newer languages such as DocBook, texinfo or schema-driven XML; or even
 ad-hoc conventions such as README files.  
-.Em Stay away from these methods!
+.Em Avoid these formats .
 Historical formats fail to capture a manual's semantic content, instead
 only modelling its style.  Newer methods requires special,
 system-specific tools and may change or become obsolete over the
 life-time of your component.
 .Pp
-There are two canonical references for writing mdoc manuals:
+There are two canonical references for writing mdoc.  Read them.
 .Pp
 .\" LIST
 .Bl -tag -width XXXXXXXXXXXXXXXX -offset indent -compact
@@ -200,10 +176,8 @@ output:
 \&.in.1:
 	mandoc -Wall,error -Tlint $<
 	cp -f $< $@
-
 \&.1.html:
 	mandoc -Thtml $< >$@
-
 \&.1.txt:
 	mandoc -Tascii $< | col -b >$@
 .Ed
@@ -238,7 +212,7 @@ symbols and so on), use the escapes dictated in
 .\" SUBSECTION
 .Ss References 
 Other components may be referenced with the
-.Sq \&Xr ,
+.Sq \&Xr
 and
 .Sq \&Sx
 macros.  Make sure that these exist.  If you intend to distribute your
@@ -266,3 +240,9 @@ lists), assume that your output device is a fixed-width terminal window:
 You may assume that the width calculated by the string literal 
 .Qq Fl o Ar outfile
 will be covered by the \-width argument.
+.\" SECTION
+.Sh MAINTENANCE
+As your component changes and bugs are fixed, your manual may become out
+of date.  You may be tempted to use automation tools like Doxygen to
+smooth the development of your manuals.  Don't.  Source documentation is
+different from a component manual.