dj.1: clarify skip/seek behavior with regards to statistics output

Makefile

@@ -16,9 +16,12 @@
 DESTDIR ?= dist
 PREFIX ?= /usr/local
 
-MANDIR != [ $(PREFIX) = / ] && printf '/usr/share/man\n' \
+# normalized prefix
+PREFIX_N != (test -d $(PREFIX) && [ '-' != $(PREFIX) ] \
+	&& CDPATH= cd -P -- $(PREFIX) && pwd -P)
+MANDIR != [ $(PREFIX_N) = / ] && printf '/usr/share/man\n' \
 	|| printf '/share/man\n'
-SYSEXITS != printf '\043include <sysexits.h>\n' | cpp -M - | sed 's/ /\n/g' \
+SYSEXITS != printf '\043include <sysexits.h>\n' | cpp -M - | tr ' ' '\n' \
 	| sed -n 's/sysexits\.h//p' || printf 'include\n'
 
 CC ?= cc
@@ -29,7 +32,7 @@ RUSTLIBS = --extern getopt=build/o/libgetopt.rlib \
 CFLAGS += -I$(SYSEXITS)
 
 .PHONY: all
-all: docs dj false fop hru intcmp mm npc rpn scrut str strcmp swab true
+all: dj false fop hru intcmp mm npc rpn scrut str strcmp swab true
 
 # keep build/include until bindgen(1) has stdin support
 # https://github.com/rust-lang/rust-bindgen/issues/2703
@@ -40,8 +43,8 @@ build:
 clean:
 	rm -rf build dist
 
-dist: all
-	mkdir -p $(DESTDIR)/$(PREFIX)/bin $(DESTDIR)/$(PREFIX)/share/man/man1
+dist: all docs
+	mkdir -p $(DESTDIR)/$(PREFIX)/bin $(DESTDIR)/$(PREFIX)/$(MANDIR)/man1
 	cp build/bin/* $(DESTDIR)/$(PREFIX)/bin
 	cp build/docs/*.1 $(DESTDIR)/$(PREFIX)/$(MANDIR)/man1
 

docs/dj.1

@@ -35,7 +35,9 @@ respectively. This language is inherited from the
 utility and used here to decrease ambiguity.
 
 When seeking or skipping to a byte, writing or reading starts at the byte
-immediately subsequent to the specified byte.
+immediately subsequent to the specified byte. Seeks and skips aren\(cqt counted
+in the output statistics because they're guaranteed to succeed (or the utility
+will exit unsuccessfully).
 .\"
 .SH OPTIONS
 

src/dj.c

@@ -16,10 +16,11 @@
  * along with this program. If not, see https://www.gnu.org/licenses/.
  */
 
+#include <assert.h> /* assert(3) */
 #include <errno.h> /* errno */
 #include <fcntl.h> /* open(2) */
 #include <stdio.h> /* fprintf(3), stderr */
-#include <stdlib.h> /* free(3), malloc(3), strtol(3), size_t */
+#include <stdlib.h> /* malloc(3), strtol(3), size_t */
 #include <string.h> /* memcpy(3), memmove(3), memset(3) */
 #include <sysexits.h> /* EX_OK, EX_USAGE */
 #include <unistd.h> /* close(2), getopt(3), lseek(2), read(2), write(2),
@@ -46,7 +47,7 @@ struct Io{
 	long seek;		/* bytes to seek/skip (will be 0 after skippage) (-sS) */
 }; 
 
-/* To be assigned to main:fmt_output and used with output(). */
+/* To be assigned to main:fmt and used with printio(). */
 static char *fmt_asv   = "%d\037%d\036%d\037%d\035%d\036%d\034";
 static char *fmt_human = "%d+%d > %d+%d; %d > %d\n";
 
@@ -58,33 +59,16 @@ static int write_flags = O_WRONLY | O_CREAT;
 
 #define MIN(a, b) (((a) < (b)) ? (a) : (b))
 
-/* Macro to check if fd is a std* file, e.g. stdin. */
-#define fdisstd(fd) \
-		((fd) == STDIN_FILENO \
-		|| (fd) == STDOUT_FILENO \
-		|| (fd) == STDERR_FILENO)
-
-/* Macro to call the cleanup functions that operate on struct io on the
- * particular io[2] used in main. Error conditions are not checked because this
- * is only used when the program is about to terminate (hence its name). */
-#define terminate(io) do{ \
-	free((io[0]).buf); \
-	free((io[1]).buf); \
-	Io_fdclose(&(io)[0]); \
-	Io_fdclose(&(io)[1]); }while(0)
-
-/* Allocates *io's buffer. Returns NULL if unsuccessful. */
-static void *
-Io_bufalloc(struct Io *io){
-
-	return (io->buf = malloc(io->bs * (sizeof *io->buf)));
-}
+/* Macro to check if fd is stdin or stdout */
+#define fdisstd(fd) ((fd) == STDIN_FILENO || (fd) == STDOUT_FILENO)
 
 /* Fills the unused portion of io's buffer with padding, updating io->bufuse.
  * Returns io. */
 static struct Io *
 Io_bufrpad(struct Io *io, int padding){
-	
+
+	assert(io != NULL);
+
 	memset(io->buf + io->bufuse, padding, io->bs - io->bufuse);
 	io->bufuse = io->bs;
 
@@ -99,6 +83,8 @@ static struct Io*
 Io_bufxapp(struct Io *dest, struct Io *src){
 	int n;
 
+	assert(dest != NULL && src != NULL);
+
 	n = MIN(src->bufuse, dest->bs - dest->bufuse);
 	memcpy(dest->buf + dest->bufuse, src->buf, n);
 	dest->bufuse += n;
@@ -115,32 +101,27 @@ Io_bufxapp(struct Io *dest, struct Io *src){
 static struct Io*
 Io_bufxfer(struct Io *dest, struct Io *src, int n){
 
+	assert(dest != NULL && src != NULL);
+
 	memcpy(dest->buf, src->buf, (dest->bufuse = n));
 	memmove(src->buf, src->buf + n, (src->bufuse -= n));
 
 	return dest;
 }
 
-/* Closes io->fn and returns -1 on error, otherwise io->fd. */
-static int
-Io_fdclose(struct Io *io){
-
-	return fdisstd(io->fd)
-		? 0
-		: close(io->fd);
-}
-
 /* Opens io->fn and saves the file descriptor into io->fd. Returns io->fd,
  * which will be -1 if an error occured. */
 static int
 Io_fdopen(struct Io *io, char *fn){
 	int fd;
-	
+
+	assert(io != NULL);
+
 	if((fd = open(fn, io->fl,
 					/* these are the flags used by touch(1p) */
 					S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP | S_IROTH | S_IWOTH))
 				!= -1
-			&& Io_fdclose(io) == 0){
+			&& (fdisstd(io->fd) || close(io->fd) == 0)){
 		io->fd = fd;
 		io->fn = fn;
 	}
@@ -148,43 +129,6 @@ Io_fdopen(struct Io *io, char *fn){
 	return fd;
 }
 
-/* Seeks io->seek bytes through *io's file descriptor, subtracting the number
- * of sought bytes from io->seek. This procedure leaves garbage in io->buf. */
-static void
-Io_fdseek(struct Io *io){
-	
-	if(io->seek != 0
-			|| (!fdisstd(io->fd) && lseek(io->fd, io->seek, SEEK_SET) != -1))
-		return;
-
-	if(io->fl == write_flags)
-		memset(io->buf, '\0', io->bs);
-
-	if(io->fl == write_flags){
-		memset(io->buf, '\0', io->bs);
-		/* We're going to cheat and use bufuse as the retval for write(2),
-		 * which is fine because it'll be zeroed as this function returns
-		 * anyway. */
-		do{
-			if((io->bufuse = write(io->fd, io->buf, MIN(io->bs, io->seek)))
-					== 0)
-				/* second chance */
-				io->bufuse = write(io->fd, io->buf, MIN(io->bs, io->seek));
-		}while((io->seek -= io->bufuse) > 0 && io->bufuse != 0);
-	}else if(io->fl == read_flags){
-		do{
-			if((io->bufuse = read(io->fd, io->buf, MIN(io->bs, io->seek)))
-					== 0)
-				/* second chance */
-				io->bufuse = read(io->fd, io->buf, MIN(io->bs, io->seek));
-		}while((io->seek -= io->bufuse) > 0 && io->bufuse != 0);
-	}
-
-	io->bufuse = 0;
-
-	return;
-}
-
 /* Reads io->bs bytes from *io's file descriptor into io->buf, storing the
  * number of read bytes in io->bufuse and updating io->bytes. If io->bufuse is
  * 0, errno will probably be set. Returns io. */
@@ -223,7 +167,7 @@ oserr(char *s){
 /* Prints statistics regarding the use of dj, particularly partially and
  * completely read and written records. */
 static void
-output(struct Io io[2], char *fmt){
+printio(char *fmt, struct Io io[2]){
 
 	fprintf(stderr, fmt,
 		io[0].rec, io[0].prec, io[1].rec, io[1].prec,
@@ -260,7 +204,7 @@ usage(char *s){
 int main(int argc, char *argv[]){
 	int align; /* low 8b used, negative if no alignment is being done */
 	int count; /* 0 if dj(1) runs until no more reads are possible */
-	char *fmt_output; /* == fmt_asv (default) or fmt_human (-H) */
+	char *fmt; /* == fmt_asv (default) or fmt_human (-H) */
 	size_t i; /* side of io being modified */
 	struct Io io[2];
 	char noerror; /* 0=exits (default) 1=retries on partial reads or writes */
@@ -268,14 +212,14 @@ int main(int argc, char *argv[]){
 	/* Set defaults. */
 	align = -1;
 	count = 0;
-	fmt_output = fmt_asv;
+	fmt = fmt_asv;
 	noerror = 0;
 	for(i = 0; i < 2; ++i){
 		io[i].bs = 1024 /* 1 KiB */; /* GNU dd(1) default; POSIX says 512B */
 		io[i].bytes = 0;
-		io[i].fd = i ? STDIN_FILENO : STDOUT_FILENO;
-		io[i].fn = i ? stdin_name   : stdout_name;
-		io[i].fl = i ? read_flags   : write_flags;
+		io[i].fd = i == 0 ? STDIN_FILENO : STDOUT_FILENO;
+		io[i].fn = i == 0 ? stdin_name   : stdout_name;
+		io[i].fl = i == 0 ? read_flags   : write_flags;
 		io[i].prec = 0;
 		io[i].rec = 0;
 		io[i].seek = 0;
@@ -285,19 +229,18 @@ int main(int argc, char *argv[]){
 		int c;
 
 		program_name = argv[0];
-		while((c = getopt(argc, argv, "a:b:B:c:i:hHns:S:o:")) != -1)
+		while((c = getopt(argc, argv, ":a:b:B:c:i:hHns:S:o:")) != -1)
 			switch(c){
 			case 'i': case 'o': i = (c == 'o');
 				if(optarg[0] == '-' && optarg[1] == '\0'){ /* optarg == "-" */
-					io[i].fd = i ? STDIN_FILENO : STDOUT_FILENO;
-					io[i].fn = i ? stdin_name   : stdout_name;
+					io[i].fd = i == 0 ? STDIN_FILENO : STDOUT_FILENO;
+					io[i].fn = i == 0 ? stdin_name   : stdout_name;
 					break;
 				}else if(Io_fdopen(&io[i], optarg) != -1)
 					break;
-				terminate(io);
 				return oserr(optarg);
 			case 'n': noerror = 1; break;
-			case 'H': fmt_output = fmt_human; break;
+			case 'H': fmt = fmt_human; break;
 			case 'a':
 				if(optarg[0] == '\0' || optarg[1] == '\0'){
 					align = optarg[0];
@@ -307,37 +250,72 @@ int main(int argc, char *argv[]){
 			case 'c': case 'b': case 's': case 'B': case 'S': /* numbers */
 				if(c == 'c' && (count = parse(optarg)) >= 0)
 					break;
-				i = (c >= 'A' && c <= 'Z'); /* uppercase changes output */
-				c &= 0x20 /* 0b 0010 0000 */; /* (ASCII) make lowercase */
+				i = (c >= 'A' && c <= 'Z');
+				c |= 0x20 /* 0b 0010 0000 */; /* (ASCII) make lowercase */
 				if((c == 'b' && (io[i].bs = parse(optarg)) > 0)
 						|| (c == 's' && (io[i].seek = parse(optarg)) >= 0))
 					break;
 				/* FALLTHROUGH */
 			default:
-				terminate(io);
 				return usage(program_name);
 			}
 	}
 
+	assert(io->fd != STDIN_FILENO  || io->fl == read_flags);
+	assert(io->fd != STDOUT_FILENO || io->fl == write_flags);
+
 	if(argc > optind){
-		terminate(io);
 		return usage(program_name);
 	}
 
 	for(i = 0; i < 2; ++i){
-		if(Io_bufalloc(&io[i]) == NULL){
+		/* buffer allocation */
+		if((io[i].buf = malloc(io[i].bs * (sizeof *(io[i].buf)))) == NULL){
 			fprintf(stderr, "%s: Failed to allocate %d bytes\n",
 				program_name, io[i].bs);
-			terminate(io);
 			return EX_OSERR;
-		}else if(io[i].seek > 0)
-			Io_fdseek(&io[i]);
-			if(io[i].seek > 0){
-				terminate(io);
-				return oserr(io[i].fn);
-			}
+		}
+		/* easy seeking */
+		if(!fdisstd(io[i].fd) && lseek(io[i].fd, io[i].seek, SEEK_SET) != -1)
+			io[i].seek = 0;
 	}
 
+	/* hard skipping */
+	if(io[0].seek > 0){
+		do{
+			if((io[0].bufuse = read(
+						io[0].fd, io[0].buf, MIN(io[0].bs, io[0].seek)))
+					== 0)
+				/* second chance */
+				io->bufuse = read(
+					io[0].fd, io[0].buf, MIN(io[0].bs, io[0].seek));
+		}while((io[0].seek -= io[0].bufuse) > 0 && io[0].bufuse != 0);
+		io[0].bufuse = 0;
+	}
+
+	/* hard seeking */
+	if(io[1].seek > 0){
+		memset(io[1].buf, '\0', io[1].bs);
+		/* We're going to cheat and use bufuse as the retval for write(2),
+		 * which is fine because it'll be zeroed as this function returns
+		 * anyway. */
+		do{
+			if((io[1].bufuse = write(
+						io[1].fd, io[1].buf, MIN(io[1].bs, io[1].seek)))
+					== 0)
+				/* second chance */
+				io[1].bufuse = write(
+					io[1].fd, io[1].buf, MIN(io[1].bs, io[1].seek));
+		}while((io[1].seek -= io[1].bufuse) > 0 && io[1].bufuse != 0);
+		io[1].bufuse = 0;
+	}
+
+	/* Sought bytes aren't counted in the statistics because successful seeking
+	 * is guaranteed here. */
+	for(i = 0; i < 2; ++i)
+		if(io[i].seek > 0)
+			return oserr(io[i].fn);
+
 	do{	/* read */
 		Io_read(&io[0]);
 		if(!noerror && io[0].bufuse == 0)
@@ -347,7 +325,7 @@ int main(int argc, char *argv[]){
 		else if(io[0].bufuse < io[0].bs){
 			++io[0].prec;
 			fprintf(stderr, "%s: Partial read:\n\t", program_name);
-			output(io, fmt_output);
+			printio(fmt, io);
 			if(!noerror)
 				count = 1;
 			if(align >= 0)
@@ -376,7 +354,7 @@ int main(int argc, char *argv[]){
 			}else if(t > io[1].bufuse && io[1].bufuse > 0){
 				io[1].prec += 1;
 				fprintf(stderr, "%s: Partial write:\n\t", program_name);
-				output(io, fmt_output);
+				printio(fmt, io);
 				if(!noerror)
 					count = 1;
 			}else if(io[1].bufuse == 0 && t < io[1].bs)
@@ -386,8 +364,7 @@ int main(int argc, char *argv[]){
 		}while(io[0].bufuse > 0);
 	}while(count == 0 || --count > 0);
 
-	output(io, fmt_output);
-	terminate(io);
+	printio(fmt, io);
 
 	return EX_OK;
 }