[PATCH 0/7] pull: whereis changes

[PATCH 0/7] pull: whereis changes

[Sami Kerola]

Hi Karel, and all,

Here are the results of the whereis(1) work which I mentioned earlier the
week in irc.  All of the patches are related to that command, and as you
can see from the diff stats a lot has changed.

The first patch is quite big.  Unfortunately I ran out of imagination how
to make it smaller, as the patch is the core change enabling rest of the
work.  Most what the 0001 has got going is with search directory list
building using new data structure.

In between patches are quite straight forward till second last which will
canonicalize output file names.  I think that is what users are
interested, but I a may be wrong as well.  Comments, anyone?

The last patch is a bug fix to RedHat item.  See commit message for
further details.

One funky side note.  After the changes the whereis seems to run about
20% quicker.  That was unintentional enhancement.


The following changes since commit 2029dbd9b168319d8a4717c1523e4facb5cff9cc:

  docs: update year (2013-03-15 15:18:22 +0100)

are available in the git repository at:

  git://github.com/kerolasa/lelux-utiliteetit.git whereis

for you to fetch changes up to 41f50bc7efcda1acc179c159ccc04e91e1ed5965:

  whereis: remove duplicates from search results (2013-03-16 23:22:02 +0000)

----------------------------------------------------------------
Sami Kerola (7):
      whereis: rewrite most of the command
      whereis: add search scope listing option
      whereis: align with howto-usage-function.txt
      whereis: support MANPATH environment variable
      docs: generic whereis.1 clean up
      whereis: canonicalize files before printing
      whereis: remove duplicates from search results

 misc-utils/Makemodule.am |   1 +
 misc-utils/whereis.1     |  93 ++++----
 misc-utils/whereis.c     | 555 ++++++++++++++++++++++++++---------------------
 3 files changed, 350 insertions(+), 299 deletions(-)

[PATCH 1/7] whereis: rewrite most of the command

[Sami Kerola]

The earlier code gave little or no change to fix bugs and improve the
command.  This rewrite attempts to make further patching easier.

Signed-off-by: Sami Kerola <kerolasa@iki.fi>
---
 misc-utils/whereis.c | 468 +++++++++++++++++++++++++--------------------------
 1 file changed, 229 insertions(+), 239 deletions(-)

diff --git a/misc-utils/whereis.c b/misc-utils/whereis.c
index 4194ad5..eeec976 100644
--- a/misc-utils/whereis.c
+++ b/misc-utils/whereis.c
@@ -49,13 +49,35 @@
 #include <stdlib.h>
 #include <string.h>
 #include <ctype.h>
+#include <assert.h>
 
 #include "xalloc.h"
 #include "nls.h"
 #include "c.h"
 #include "closestream.h"
 
-static char *bindirs[] = {
+#define BIN_BIT (1 << 1)
+#define MAN_BIT (1 << 2)
+#define SRC_BIT (1 << 3)
+
+static char uflag = 0;
+
+enum {
+	BIN_DIR = 0,
+	MAN_DIR,
+	SRC_DIR,
+	ALL_DIRS
+};
+
+struct is_here {
+	int type;
+	dev_t st_dev;
+	ino_t st_ino;
+	char *path;
+	struct is_here *next;
+};
+
+static const char *bindirs[] = {
 	"/bin",
 	"/usr/bin",
 	"/sbin",
@@ -110,7 +132,7 @@ static char *bindirs[] = {
 	0
 };
 
-static char *mandirs[] = {
+static const char *mandirs[] = {
 	"/usr/man/*",
 	"/usr/share/man/*",
 	"/usr/X386/man/*",
@@ -120,7 +142,7 @@ static char *mandirs[] = {
 	0
 };
 
-static char *srcdirs[] = {
+static const char *srcdirs[] = {
 	"/usr/src/*",
 	"/usr/src/lib/libc/*",
 	"/usr/src/lib/libc/net/*",
@@ -130,11 +152,7 @@ static char *srcdirs[] = {
 	0
 };
 
-static char sflag = 1, bflag = 1, mflag = 1, uflag;
-static char **Sflag, **Bflag, **Mflag, **pathdir, **pathdir_p;
-static int Scnt, Bcnt, Mcnt, count, print;
-
-static void __attribute__ ((__noreturn__)) usage(FILE * out)
+static void __attribute__((__noreturn__)) usage(FILE *out)
 {
 	fputs(_("\nUsage:\n"), out);
 	fprintf(out,
@@ -156,8 +174,7 @@ static void __attribute__ ((__noreturn__)) usage(FILE * out)
 	exit(out == stderr ? EXIT_FAILURE : EXIT_SUCCESS);
 }
 
-static int
-itsit(char *cp, char *dp)
+static int itsit(char *cp, char *dp)
 {
 	int i = strlen(dp);
 
@@ -185,86 +202,90 @@ itsit(char *cp, char *dp)
 	return 0;
 }
 
-static void
-findin(char *dir, char *cp)
+static void findin(char *dir, char *cp, int *nr_found, char **wait)
 {
 	DIR *dirp;
 	struct dirent *dp;
-	char *d, *dd;
-	size_t l;
-	char dirbuf[1024];
-	struct stat statbuf;
 
-	dd = strchr(dir, '*');
-	if (!dd) {
-		dirp = opendir(dir);
-		if (dirp == NULL)
-			return;
-		while ((dp = readdir(dirp)) != NULL) {
-			if (itsit(cp, dp->d_name)) {
-				count++;
-				if (print)
-					printf(" %s/%s", dir, dp->d_name);
-			}
-		}
-		closedir(dirp);
+	dirp = opendir(dir);
+	if (dirp == NULL)
 		return;
-	}
-
-	l = strlen(dir);
-	if (l < sizeof(dirbuf)) {
-		/* refuse excessively long names */
-		strcpy(dirbuf, dir);
-		d = strchr(dirbuf, '*');
-		if (d)
-			*d = 0;
-		dirp = opendir(dirbuf);
-		if (dirp == NULL)
-			return;
-		while ((dp = readdir(dirp)) != NULL) {
-			if (!strcmp(dp->d_name, ".") ||
-			    !strcmp(dp->d_name, ".."))
-				continue;
-			if (strlen(dp->d_name) + l > sizeof(dirbuf))
-				continue;
-			sprintf(d, "%s", dp->d_name);
-			if (stat(dirbuf, &statbuf))
-				continue;
-			if (!S_ISDIR(statbuf.st_mode))
-				continue;
-			strcat(d, dd + 1);
-			findin(dirbuf, cp);
+	while ((dp = readdir(dirp)) != NULL)
+		if (itsit(cp, dp->d_name)) {
+			if (*(nr_found) == 0)
+				xasprintf(wait, "%s/%s", dir, dp->d_name);
+			else if (*(nr_found) == 1 && *wait != NULL)
+				printf("%s: %s %s/%s", cp, *wait, dir,
+				       dp->d_name);
+			else if (*(nr_found) == 1)
+				printf("%s: %s/%s", cp, dir, dp->d_name);
+			else
+				printf(" %s/%s", dir, dp->d_name);
+			(*nr_found)++;
 		}
-		closedir(dirp);
-	}
+	closedir(dirp);
 	return;
-
 }
 
-static int inpath(const char *str)
+static void add_dir(struct is_here *dirlist, int type, const char *dir)
 {
-	size_t i;
+	struct stat statbuf;
+	struct is_here *prev = NULL;
 
-	for (i = 0; i < ARRAY_SIZE(bindirs) - 1 ; i++)
-		if (!strcmp(bindirs[i], str))
-			return 1;
+	if (access(dir, R_OK) != 0)
+		return;
+	if (stat(dir, &statbuf))
+		return;
+	if (!S_ISDIR(statbuf.st_mode))
+		return;
+	while (dirlist) {
+		if (dirlist->st_ino == statbuf.st_ino &&
+		    dirlist->st_dev == statbuf.st_dev && dirlist->type == type)
+			return;
+		prev = dirlist;
+		dirlist = dirlist->next;
+	}
+	assert(prev);
+	dirlist = xcalloc(1, sizeof(struct is_here));
+	dirlist->st_ino = statbuf.st_ino;
+	dirlist->st_dev = statbuf.st_dev;
+	dirlist->type = type;
+	dirlist->path = xstrdup(dir);
+	prev->next = dirlist;
+	return;
+}
 
-	for (i = 0; i < ARRAY_SIZE(mandirs) - 1; i++)
-		if (!strcmp(mandirs[i], str))
-			return 1;
+static void add_subdirs(struct is_here *dirlist, int type,
+			const char *dir)
+{
+#define DIRBUF 1024
+	char dirbuf[DIRBUF], *d;
+	DIR *dirp;
+	struct dirent *dp;
 
-	for (i = 0; i < ARRAY_SIZE(srcdirs) - 1; i++)
-		if (!strcmp(srcdirs[i], str))
-			return 1;
+	strncpy(dirbuf, dir, DIRBUF);
+	d = strchr(dirbuf, '*');
+	*d = 0;
 
-	return 0;
+	dirp = opendir(dirbuf);
+	if (dirp == NULL)
+		return;
+	while ((dp = readdir(dirp)) != NULL) {
+		if (!strcmp(dp->d_name, ".") || !strcmp(dp->d_name, ".."))
+			continue;
+		sprintf(d, "%s", dp->d_name);
+		/* a dir definition can have a star in middle of path */
+		strcat(dirbuf, strchr(dir, '*') + 1);
+		add_dir(dirlist, type, dirbuf);
+	}
+	closedir(dirp);
+	return;
 }
 
-static void fillpath(void)
+static void environ_list(const char *env, struct is_here *dirlist,
+			 int type)
 {
-	char *key=NULL, *tok=NULL, *pathcp, *path = getenv("PATH");
-	int i = 0;
-
+	char *key = NULL, *tok = NULL, *pathcp, *path = getenv(env);
 
 	if (!path)
 		return;
@@ -272,150 +293,105 @@ static void fillpath(void)
 
 	for (tok = strtok_r(pathcp, ":", &key); tok;
 	     tok = strtok_r(NULL, ":", &key)) {
-
-		/* make sure we don't repeat the search path */
-		if (inpath(tok))
-			continue;
-
-		pathdir = xrealloc(pathdir, (i + 1) * sizeof(char *));
-		pathdir[i++] = xstrdup(tok);
+		add_dir(dirlist, type, tok);
 	}
-
-	pathdir = xrealloc(pathdir, (i + 1) * sizeof(char *));
-	pathdir[i] = NULL;
-
-	pathdir_p = pathdir;
 	free(pathcp);
+	return;
 }
 
-static void freepath(void)
-{
-	free(pathdir);
-}
-
-static void
-findv(char **dirv, int dirc, char *cp)
-{
-
-	while (dirc > 0)
-		findin(*dirv++, cp), dirc--;
-}
-
-static void
-looksrc(char *cp)
-{
-	if (Sflag == NULL)
-		findv(srcdirs, ARRAY_SIZE(srcdirs)-1, cp);
-	else
-		findv(Sflag, Scnt, cp);
-}
-
-static void
-lookbin(char *cp)
-{
-	if (Bflag == NULL) {
-		findv(bindirs, ARRAY_SIZE(bindirs)-1, cp);
-		while (*pathdir_p)
-			findin(*pathdir_p++, cp);		/* look $PATH */
-	 } else
-		findv(Bflag, Bcnt, cp);
-}
-
-static void
-lookman(char *cp)
+static void construct_list(struct is_here *dirlist, int type,
+			   const char **list)
 {
-	if (Mflag == NULL)
-		findv(mandirs, ARRAY_SIZE(mandirs)-1, cp);
-	else
-		findv(Mflag, Mcnt, cp);
+	size_t i;
+	char *star;
+	for (i = 0; list[i]; i++) {
+		star = strchr(list[i], '*');
+		if (!star)
+			add_dir(dirlist, type, list[i]);
+		else
+			add_subdirs(dirlist, type, list[i]);
+	}
+	return;
 }
 
-static void
-getlist(int *argcp, char ***argvp, char ***flagp, int *cntp)
+static void getlist(int *argcp, char ***argvp, struct is_here *dirlist,
+		   int type)
 {
 	(*argvp)++;
-	*flagp = *argvp;
-	*cntp = 0;
-	for ((*argcp)--; *argcp > 0 && (*argvp)[0][0] != '-'; (*argcp)--)
-		(*cntp)++, (*argvp)++;
-	(*argcp)++;
-	(*argvp)--;
+	for ((*argcp)--; 0 < *argcp && (*argvp)[0][0] != '-'; (*argcp)--) {
+		add_dir(dirlist, type, **argvp);
+		(*argvp)++;
+	}
+	return;
 }
 
-static void
-zerof(void)
+static int check_type(int type, int flags)
 {
-	if (sflag && bflag && mflag)
-		sflag = bflag = mflag = 0;
+	if (type == BIN_DIR && !(flags & BIN_BIT))
+		return 1;
+	if (type == MAN_DIR && !(flags & MAN_BIT))
+		return 1;
+	if (type == SRC_DIR && !(flags & SRC_BIT))
+		return 1;
+	return 0;
 }
 
-static int
-print_again(char *cp)
+static struct is_here *free_dirs(struct is_here *dirlist, int type)
 {
-	if (print)
-		printf("%s:", cp);
-	if (sflag) {
-		looksrc(cp);
-		if (uflag && print == 0 && count != 1) {
-			print = 1;
-			return 1;
+	struct is_here *first, *prev, *next;
+	if (type == ALL_DIRS) {
+		while (dirlist) {
+			next = dirlist->next;
+			free(dirlist->path);
+			free(dirlist);
+			dirlist = next;
 		}
+		return NULL;
 	}
-	count = 0;
-	if (bflag) {
-		lookbin(cp);
-		if (uflag && print == 0 && count != 1) {
-			print = 1;
-			return 1;
-		}
-	}
-	count = 0;
-	if (mflag) {
-		lookman(cp);
-		if (uflag && print == 0 && count != 1) {
-			print = 1;
-			return 1;
+	first = NULL;
+	prev = NULL;
+	while (dirlist) {
+		if (dirlist->type == type) {
+			next = dirlist->next;
+			free(dirlist->path);
+			free(dirlist);
+			dirlist = next;
+			if (prev)
+				prev->next = dirlist;
+			continue;
 		}
+		if (!first)
+			first = dirlist;
+		prev = dirlist;
+		dirlist = dirlist->next;
 	}
-	return 0;
+	return first;
 }
 
-static void
-lookup(char *cp)
+static void lookup(char *argv, struct is_here *dirlist, int flags)
 {
-	register char *dp;
-
-	for (dp = cp; *dp; dp++)
-		continue;
-	for (; dp > cp; dp--) {
-		if (*dp == '.') {
-			*dp = 0;
-			break;
-		}
+	int nr_found;
+	char *wait = NULL;
+
+	nr_found = uflag == 1 ? 0 : 1;
+
+	for (; dirlist; dirlist = dirlist->next) {
+		if (check_type(dirlist->type, flags))
+			continue;
+		if (dirlist->path != NULL)
+			findin(dirlist->path, argv, &nr_found, &wait);
 	}
-	for (dp = cp; *dp; dp++)
-		if (*dp == '/')
-			cp = dp + 1;
-	if (uflag) {
-		print = 0;
-		count = 0;
-	} else
-		print = 1;
-
-	while (print_again(cp))
-		/* all in print_again() */ ;
-
-	if (print)
-		printf("\n");
+	free(wait);
+	if (1 < nr_found)
+		putchar('\n');
+	return;
 }
 
-/*
- * whereis name
- * look for source, documentation and binaries
- */
-int
-main(int argc, char **argv)
+int main(int argc, char **argv)
 {
+	struct is_here *dirlist;
+	char flags = CHAR_MAX;
+
 	setlocale(LC_ALL, "");
 	bindtextdomain(PACKAGE, LOCALEDIR);
 	textdomain(PACKAGE);
@@ -425,60 +401,74 @@ main(int argc, char **argv)
 	if (argc == 0)
 		usage(stderr);
 
+	dirlist = xcalloc(1, sizeof(struct is_here));
+	construct_list(dirlist, BIN_DIR, bindirs);
+	environ_list("PATH", dirlist, BIN_DIR);
+	construct_list(dirlist, MAN_DIR, mandirs);
+	construct_list(dirlist, SRC_DIR, srcdirs);
+
 	do
 		if (argv[0][0] == '-') {
 			register char *cp = argv[0] + 1;
-			while (*cp) switch (*cp++) {
-
-			case 'f':
-				break;
-
-			case 'S':
-				getlist(&argc, &argv, &Sflag, &Scnt);
-				break;
-
-			case 'B':
-				getlist(&argc, &argv, &Bflag, &Bcnt);
-				break;
-
-			case 'M':
-				getlist(&argc, &argv, &Mflag, &Mcnt);
-				break;
-
-			case 's':
-				zerof();
-				sflag++;
-				continue;
-
-			case 'u':
-				uflag++;
-				continue;
-
-			case 'b':
-				zerof();
-				bflag++;
-				continue;
-
-			case 'm':
-				zerof();
-				mflag++;
-				continue;
-			case 'V':
-				printf(UTIL_LINUX_VERSION);
-				return EXIT_SUCCESS;
-			case 'h':
-				usage(stdout);
-			default:
-				usage(stderr);
-			}
+			while (*cp)
+				switch (*cp++) {
+
+				case 'f':
+					break;
+
+				case 'u':
+					uflag = 1;
+					continue;
+
+				case 'B':
+					dirlist = free_dirs(dirlist, BIN_DIR);
+					getlist(&argc, &argv, dirlist, BIN_DIR);
+					break;
+
+				case 'M':
+					dirlist = free_dirs(dirlist, MAN_DIR);
+					getlist(&argc, &argv, dirlist, MAN_DIR);
+					break;
+
+				case 'S':
+					dirlist = free_dirs(dirlist, SRC_DIR);
+					getlist(&argc, &argv, dirlist, SRC_DIR);
+					break;
+
+				case 'b':
+					if (flags == CHAR_MAX)
+						flags = BIN_BIT;
+					else
+						flags |= BIN_BIT;
+					continue;
+
+				case 'm':
+					if (flags == CHAR_MAX)
+						flags = MAN_BIT;
+					else
+						flags |= MAN_BIT;
+					continue;
+
+				case 's':
+					if (flags == CHAR_MAX)
+						flags = SRC_BIT;
+					else
+						flags |= SRC_BIT;
+					continue;
+
+				case 'V':
+					printf(UTIL_LINUX_VERSION);
+					return EXIT_SUCCESS;
+				case 'h':
+					usage(stdout);
+				default:
+					usage(stderr);
+				}
 			argv++;
 		} else {
-			if (Bcnt == 0 && pathdir == NULL)
-				fillpath();
-			lookup(*argv++);
+			lookup(*argv++, dirlist, flags);
 		}
 	while (--argc > 0);
-
-	freepath();
+	free_dirs(dirlist, ALL_DIRS);
 	return EXIT_SUCCESS;
 }
-- 
1.8.2

[PATCH 2/7] whereis: add search scope listing option

[Sami Kerola]

Mostly useful when debugging why the command does, or does not, work.

Signed-off-by: Sami Kerola <kerolasa@iki.fi>
---
 misc-utils/whereis.1 | 11 ++++++++++-
 misc-utils/whereis.c | 28 ++++++++++++++++++++++++++++
 2 files changed, 38 insertions(+), 1 deletion(-)

diff --git a/misc-utils/whereis.1 b/misc-utils/whereis.1
index 86e6a39..9b33999 100644
--- a/misc-utils/whereis.1
+++ b/misc-utils/whereis.1
@@ -91,7 +91,16 @@ be used when any of the
 or
 .BR \-S
 options is used.
-
+.IP "\fB\-l"
+Output list of effective lookup paths the
+.B whereis
+is using.  When non of
+.BR \-B ,
+.BR \-M ,
+or
+.BR \-S
+is specified the option will out hard coded paths that the command was
+able to find on system.
 .SH EXAMPLE
 To find all files in
 .B /usr/bin
diff --git a/misc-utils/whereis.c b/misc-utils/whereis.c
index eeec976..f579e29 100644
--- a/misc-utils/whereis.c
+++ b/misc-utils/whereis.c
@@ -167,6 +167,7 @@ static void __attribute__((__noreturn__)) usage(FILE *out)
 		" -s         search only for sources\n"
 		" -S <dirs>  define sources lookup path\n"
 		" -u         search for unusual entries\n"
+		" -l         output effective lookup paths\n"
 		" -V         output version information and exit\n"
 		" -h         display this help and exit\n\n"), out);
 
@@ -387,6 +388,29 @@ static void lookup(char *argv, struct is_here *dirlist, int flags)
 	return;
 }
 
+static void list_search_dir(struct is_here *dirlist)
+{
+	while (dirlist) {
+		if (dirlist->path != NULL) {
+			switch (dirlist->type) {
+			case BIN_DIR:
+				printf("bin: ");
+				break;
+			case MAN_DIR:
+				printf("man: ");
+				break;
+			case SRC_DIR:
+				printf("src: ");
+				break;
+			default:
+				abort();
+			}
+			printf("%s\n", dirlist->path);
+		}
+		dirlist = dirlist->next;
+	}
+}
+
 int main(int argc, char **argv)
 {
 	struct is_here *dirlist;
@@ -456,6 +480,10 @@ int main(int argc, char **argv)
 						flags |= SRC_BIT;
 					continue;
 
+				case 'l':
+					list_search_dir(dirlist);
+					continue;
+
 				case 'V':
 					printf(UTIL_LINUX_VERSION);
 					return EXIT_SUCCESS;
-- 
1.8.2

[PATCH 3/7] whereis: align with howto-usage-function.txt

[Sami Kerola]

Signed-off-by: Sami Kerola <kerolasa@iki.fi>
---
 misc-utils/whereis.c | 33 +++++++++++++++------------------
 1 file changed, 15 insertions(+), 18 deletions(-)

diff --git a/misc-utils/whereis.c b/misc-utils/whereis.c
index f579e29..e3cfb5f 100644
--- a/misc-utils/whereis.c
+++ b/misc-utils/whereis.c
@@ -154,24 +154,21 @@ static const char *srcdirs[] = {
 
 static void __attribute__((__noreturn__)) usage(FILE *out)
 {
-	fputs(_("\nUsage:\n"), out);
-	fprintf(out,
-	      _(" %s [options] file\n"), program_invocation_short_name);
-
-	fputs(_("\nOptions:\n"), out);
-	fputs(_(" -f <file>  define search scope\n"
-		" -b         search only for binaries\n"
-		" -B <dirs>  define binaries lookup path\n"
-		" -m         search only for manuals\n"
-		" -M <dirs>  define man lookup path\n"
-		" -s         search only for sources\n"
-		" -S <dirs>  define sources lookup path\n"
-		" -u         search for unusual entries\n"
-		" -l         output effective lookup paths\n"
-		" -V         output version information and exit\n"
-		" -h         display this help and exit\n\n"), out);
-
-	fputs(_("See how to use file and dirs arguments from whereis(1) manual.\n"), out);
+	fputs(USAGE_HEADER, out);
+	fprintf(out, _(" %s [options] file\n"), program_invocation_short_name);
+
+	fputs(USAGE_OPTIONS, out);
+	fputs(_(" -b         search only for binaries\n"), out);
+	fputs(_(" -B <dirs>  define binaries lookup path\n"), out);
+	fputs(_(" -m         search only for manuals\n"), out);
+	fputs(_(" -M <dirs>  define man lookup path\n"), out);
+	fputs(_(" -s         search only for sources\n"), out);
+	fputs(_(" -S <dirs>  define sources lookup path\n"), out);
+	fputs(_(" -f         terminate <dirs> argument list\n"), out);
+	fputs(_(" -u         search for unusual entries\n"), out);
+	fputs(_(" -l         output effective lookup paths\n"), out);
+	fprintf(out, USAGE_MAN_TAIL("whereis(1)"));
+
 	exit(out == stderr ? EXIT_FAILURE : EXIT_SUCCESS);
 }
 
-- 
1.8.2

[PATCH 4/7] whereis: support MANPATH environment variable

[Sami Kerola]

The MANPATH has been around long enough that it can be added to whereis
default search path.

Reference: manpath(1)
Addresses: http://pubs.opengroup.org/onlinepubs/7908799/xbd/envvar.html
Signed-off-by: Sami Kerola <kerolasa@iki.fi>
---
 misc-utils/whereis.1 | 31 +++++++++++++++++--------------
 misc-utils/whereis.c |  1 +
 2 files changed, 18 insertions(+), 14 deletions(-)

diff --git a/misc-utils/whereis.1 b/misc-utils/whereis.1
index 9b33999..4ce7a07 100644
--- a/misc-utils/whereis.1
+++ b/misc-utils/whereis.1
@@ -30,7 +30,7 @@
 .\" SUCH DAMAGE.
 .\"
 .\" @(#)whereis.1 from UCB 4.2
-.TH WHEREIS 1 "January 2013" "util-linux" "User Commands"
+.TH WHEREIS 1 "March 2013" "util-linux" "User Commands"
 .SH NAME
 whereis \- locate the binary, source, and manual page files for a command
 .SH SYNOPSIS
@@ -113,21 +113,24 @@ or have no source in
 .B $ cd /usr/bin
 .br
 .B $ whereis \-u \-ms \-M /usr/man/man1 \-S /usr/src \-f *
-
-.SH FILES
+.SH "FILE SEARCH PATHS"
+By default
 .B whereis
-contains a basic set of hard-coded paths (see below), but if the
-.B \-B
-option is not specified, it also follows the
+tries to find files from hard-coded paths, which are defined with glob
+patterns. The command attempst to use contents of
 .B $PATH
-environment variable (since version 2.21).
-.IP
-.br
-/{bin,sbin,etc}
-.br
-/usr/{lib,\:bin,\:old,\:new,\:local,\:games,\:include,\:etc,\:src,\:man,\:sbin,\:X386,\:TeX,\:g++-include}
-.br
-/usr/local/{X386,\:TeX,\:X11,\:include,\:lib,\:man,\:etc,\:bin,\:games,\:emacs}
+and
+.B $MANPATH
+environment variables as default search path.  The easiest way to know
+what paths are in use is to add
+.B \-l
+listing option.  Effects of the
+.BR \-B ,
+.BR \-M ,
+and
+.BR \-S
+are display with
+.BR \-l .
 .SH "SEE ALSO"
 .BR chdir (2)
 .SH BUGS
diff --git a/misc-utils/whereis.c b/misc-utils/whereis.c
index e3cfb5f..3fad843 100644
--- a/misc-utils/whereis.c
+++ b/misc-utils/whereis.c
@@ -426,6 +426,7 @@ int main(int argc, char **argv)
 	construct_list(dirlist, BIN_DIR, bindirs);
 	environ_list("PATH", dirlist, BIN_DIR);
 	construct_list(dirlist, MAN_DIR, mandirs);
+	environ_list("MANPATH", dirlist, MAN_DIR);
 	construct_list(dirlist, SRC_DIR, srcdirs);
 
 	do
-- 
1.8.2

[PATCH 5/7] docs: generic whereis.1 clean up

[Sami Kerola]

Align with howto-man-page.txt, and more importantly remove out dated 'SEE
ALSO' and 'BUGS' sections.

Signed-off-by: Sami Kerola <kerolasa@iki.fi>
---
 misc-utils/whereis.1 | 61 +++++++++++++++++++++-------------------------------
 1 file changed, 25 insertions(+), 36 deletions(-)

diff --git a/misc-utils/whereis.1 b/misc-utils/whereis.1
index 4ce7a07..fde0dd7 100644
--- a/misc-utils/whereis.1
+++ b/misc-utils/whereis.1
@@ -42,8 +42,8 @@ whereis \- locate the binary, source, and manual page files for a command
 .SH DESCRIPTION
 .B whereis
 locates the binary, source and manual files for the specified command names.
-The supplied names are first stripped of leading pathname components
-and any (single) trailing extension of the form
+The supplied names are first stripped of leading pathname components and any
+(single) trailing extension of the form
 .BI . ext
 (for example:
 .BR .c )
@@ -51,25 +51,26 @@ Prefixes of
 .B s.
 resulting from use of source code control are also dealt with.
 .B whereis
-then attempts to locate the desired program in
-the standard Linux places, and in the places specified by
-.BR $PATH .
-
+then attempts to locate the desired program in the standard Linux places, and
+in the places specified by
+.B $PATH
+and
+.BR $MANPATH .
 .SH OPTIONS
 .TP
-.IP "\fB\-b\fP"
+.IP \fB\-b\fP
 Search only for binaries.
-.IP "\fB\-m\fP"
+.IP \fB\-m\fP
 Search only for manuals.
-.IP "\fB\-s\fP"
+.IP \fB\-s\fP
 Search only for sources.
-.IP "\fB\-u\fP"
-Only show the command names that have unusual entries.  A command
-is said to be unusual if it does
-not have just one entry of each explicitly requested type.  Thus
-.RB ` "whereis\ \ \-m\ \ \-u\ \ *" '
-asks for those files in the current
-directory which have no documentation file, or more than one.
+.IP \fB\-u\fP
+Only show the command names that have unusual entries.  A command is said to be
+unusual if it does not have just one entry of each explicitly requested type.
+Thus
+.RB ' "whereis \-m \-u *" '
+asks for those files in the current directory which have no documentation file,
+or more than one.
 .IP "\fB\-B \fIlist\fP"
 Limit the places where
 .B whereis
@@ -99,16 +100,16 @@ is using.  When non of
 .BR \-M ,
 or
 .BR \-S
-is specified the option will out hard coded paths that the command was
-able to find on system.
+is specified the option will out hard coded paths that the command was able to
+find on system.
 .SH EXAMPLE
 To find all files in
-.B /usr/bin
+.B /usr/\:bin
 which are not documented
 in
-.B /usr/man/man1
+.B /usr/\:man/\:man1
 or have no source in
-.BR /usr/src :
+.BR /usr/\:src :
 .IP
 .B $ cd /usr/bin
 .br
@@ -131,21 +132,9 @@ and
 .BR \-S
 are display with
 .BR \-l .
-.SH "SEE ALSO"
-.BR chdir (2)
-.SH BUGS
-Since
-.B whereis
-uses
-.BR chdir (2)
-to run faster, pathnames given with the
-.BR \-B ,
-.BR \-M ,
-or
-.BR \-S
-options must be full; that is, they must begin with a
-.RB ` / '.
 .PP
 .SH AVAILABILITY
 The whereis command is part of the util-linux package and is available from
-ftp://ftp.kernel.org/pub/linux/utils/util-linux/.
+.UR ftp://\:ftp.kernel.org\:/pub\:/linux\:/utils\:/util-linux/
+Linux Kernel Archive
+.UE .
-- 
1.8.2

[PATCH 6/7] whereis: canonicalize files before printing

[Sami Kerola]

The whereis gives promise to find files so it seems to be more correct
print symlink destinations than locations of symlinks.

Signed-off-by: Sami Kerola <kerolasa@iki.fi>
---
 misc-utils/Makemodule.am |  1 +
 misc-utils/whereis.c     | 20 +++++++++++++-------
 2 files changed, 14 insertions(+), 7 deletions(-)

diff --git a/misc-utils/Makemodule.am b/misc-utils/Makemodule.am
index b77c30a..a615047 100644
--- a/misc-utils/Makemodule.am
+++ b/misc-utils/Makemodule.am
@@ -44,6 +44,7 @@ namei_SOURCES = misc-utils/namei.c lib/strutils.c
 usrbin_exec_PROGRAMS += whereis
 dist_man_MANS += misc-utils/whereis.1
 whereis_SOURCES = misc-utils/whereis.c
+whereis_LDADD = $(LDADD) libcommon.la
 
 if BUILD_LSLOCKS
 usrbin_exec_PROGRAMS += lslocks
diff --git a/misc-utils/whereis.c b/misc-utils/whereis.c
index 3fad843..5136eff 100644
--- a/misc-utils/whereis.c
+++ b/misc-utils/whereis.c
@@ -55,6 +55,7 @@
 #include "nls.h"
 #include "c.h"
 #include "closestream.h"
+#include "canonicalize.h"
 
 #define BIN_BIT (1 << 1)
 #define MAN_BIT (1 << 2)
@@ -204,21 +205,26 @@ static void findin(char *dir, char *cp, int *nr_found, char **wait)
 {
 	DIR *dirp;
 	struct dirent *dp;
+	char *full_path, *rl;
 
 	dirp = opendir(dir);
 	if (dirp == NULL)
 		return;
 	while ((dp = readdir(dirp)) != NULL)
 		if (itsit(cp, dp->d_name)) {
+			xasprintf(&full_path, "%s/%s", dir, dp->d_name);
+			rl = canonicalize_path(full_path);
+			free(full_path);
 			if (*(nr_found) == 0)
-				xasprintf(wait, "%s/%s", dir, dp->d_name);
-			else if (*(nr_found) == 1 && *wait != NULL)
-				printf("%s: %s %s/%s", cp, *wait, dir,
-				       dp->d_name);
-			else if (*(nr_found) == 1)
-				printf("%s: %s/%s", cp, dir, dp->d_name);
+				wait = &rl;
+			else if (*(nr_found) == 1 && *wait != NULL) {
+				printf("%s: %s %s", cp, *wait, rl);
+				free(*wait);
+			} else if (*(nr_found) == 1)
+				printf("%s: %s", cp, rl);
 			else
-				printf(" %s/%s", dir, dp->d_name);
+				printf(" %s", rl);
+			free(rl);
 			(*nr_found)++;
 		}
 	closedir(dirp);
-- 
1.8.2

[PATCH 7/7] whereis: remove duplicates from search results

[Sami Kerola]

Now when the whereis will not have search directory duplicates, and it
finds real path for symlinked files in paths it is time to remove search
result duplicates.

Addresses: https://bugzilla.redhat.com/show_bug.cgi?id=835211
Signed-off-by: Sami Kerola <kerolasa@iki.fi>
---
 misc-utils/whereis.c | 31 +++++++++++++++++++++++++++++--
 1 file changed, 29 insertions(+), 2 deletions(-)

diff --git a/misc-utils/whereis.c b/misc-utils/whereis.c
index 5136eff..0e333e1 100644
--- a/misc-utils/whereis.c
+++ b/misc-utils/whereis.c
@@ -173,6 +173,28 @@ static void __attribute__((__noreturn__)) usage(FILE *out)
 	exit(out == stderr ? EXIT_FAILURE : EXIT_SUCCESS);
 }
 
+static int is_same_file(char *file, struct is_here *filelist)
+{
+	struct stat statbuf;
+	struct is_here *prev = NULL;
+
+	if (stat(file, &statbuf))
+		return 1;
+	while (filelist) {
+		if (filelist->st_ino == statbuf.st_ino &&
+		    filelist->st_dev == statbuf.st_dev)
+			return 1;
+		prev = filelist;
+		filelist = filelist->next;
+	}
+	assert(prev);
+	filelist = xcalloc(1, sizeof(struct is_here));
+	filelist->st_ino = statbuf.st_ino;
+	filelist->st_dev = statbuf.st_dev;
+	prev->next = filelist;
+	return 0;
+}
+
 static int itsit(char *cp, char *dp)
 {
 	int i = strlen(dp);
@@ -201,7 +223,7 @@ static int itsit(char *cp, char *dp)
 	return 0;
 }
 
-static void findin(char *dir, char *cp, int *nr_found, char **wait)
+static void findin(char *dir, char *cp, int *nr_found, char **wait, struct is_here *filelist)
 {
 	DIR *dirp;
 	struct dirent *dp;
@@ -213,6 +235,8 @@ static void findin(char *dir, char *cp, int *nr_found, char **wait)
 	while ((dp = readdir(dirp)) != NULL)
 		if (itsit(cp, dp->d_name)) {
 			xasprintf(&full_path, "%s/%s", dir, dp->d_name);
+			if (is_same_file(full_path, filelist))
+				continue;
 			rl = canonicalize_path(full_path);
 			free(full_path);
 			if (*(nr_found) == 0)
@@ -376,16 +400,19 @@ static void lookup(char *argv, struct is_here *dirlist, int flags)
 {
 	int nr_found;
 	char *wait = NULL;
+	struct is_here *filelist;
 
 	nr_found = uflag == 1 ? 0 : 1;
+	filelist = xcalloc(1, sizeof(struct is_here));
 
 	for (; dirlist; dirlist = dirlist->next) {
 		if (check_type(dirlist->type, flags))
 			continue;
 		if (dirlist->path != NULL)
-			findin(dirlist->path, argv, &nr_found, &wait);
+			findin(dirlist->path, argv, &nr_found, &wait, filelist);
 	}
 	free(wait);
+	free_dirs(filelist, ALL_DIRS);
 	if (1 < nr_found)
 		putchar('\n');
 	return;
-- 
1.8.2

Re: [PATCH 0/7] pull: whereis changes

[Sami Kerola]

On Sat, Mar 16, 2013 at 11:40 PM, Sami Kerola <kerolasa@iki.fi> wrote:
> Here are the results of the whereis(1) work which I mentioned earlier the
> week in irc.  All of the patches are related to that command, and as you
> can see from the diff stats a lot has changed.
>
> The first patch is quite big.  Unfortunately I ran out of imagination how
> to make it smaller, as the patch is the core change enabling rest of the
> work.  Most what the 0001 has got going is with search directory list
> building using new data structure.
>
> In between patches are quite straight forward till second last which will
> canonicalize output file names.  I think that is what users are
> interested, but I a may be wrong as well.  Comments, anyone?
>
> The last patch is a bug fix to RedHat item.  See commit message for
> further details.
>
> One funky side note.  After the changes the whereis seems to run about
> 20% quicker.  That was unintentional enhancement.

Hi,

Is this regression, or a command behavior change resulting to correct
output when not patched? I think earlier whereis misbehaved. Reason
being is that shell 'which' or path completion does not recognize
broken symlink.

$ cd ~/bin
$ ln -s bar foo
$ whereis foo
foo: /home/kerolasa/bin/foo
$ cd ~/src/util-linux
$ ./whereis foo
# no results
$ git diff
diff --git a/misc-utils/whereis.c b/misc-utils/whereis.c
index 0e333e1..30a6819 100644
--- a/misc-utils/whereis.c
+++ b/misc-utils/whereis.c
@@ -179,7 +179,8 @@ static int is_same_file(char *file, struct is_here
*filelist)
        struct is_here *prev = NULL;

        if (stat(file, &statbuf))
-               return 1;
+               if (lstat(file, &statbuf))
+                       return 1;
        while (filelist) {
                if (filelist->st_ino == statbuf.st_ino &&
                    filelist->st_dev == statbuf.st_dev)
$ make
# <...>
./whereis foo
foo: /home/kerolasa/bin/foo

-- 
   Sami Kerola
   http://www.iki.fi/kerolasa/

Re: [PATCH 6/7] whereis: canonicalize files before printing

[Karel Zak]

On Sat, Mar 16, 2013 at 11:40:18PM +0000, Sami Kerola wrote:
> The whereis gives promise to find files so it seems to be more correct
> print symlink destinations than locations of symlinks.

 Not sure. I think it's correct to canonicalize directory paths (to
 avoid problems like /bin -> /usr/bin symlinks), but I don't see
 a reason to canonicalize whole paths (filenames).

 For example

    $ whereis -b bzcat
    bzcat: /usr/bin/bzcat

 is expected answer, but

    $ whereis -b bzcat
    bzcat: /usr/bin/bzip2

 looks strange. The another story is that many things are redirected
 by symlinks to /etc/alternatives and then to pretty specific places.

 Maybe we can add an option to canonicalize, but from my point of view 
 it should not be enabled by default.

 I have applied patch to canonicalize directory names.

    Karel

-- 
 Karel Zak  <kzak@redhat.com>
 http://karelzak.blogspot.com

Re: [PATCH 7/7] whereis: remove duplicates from search results

[Karel Zak]

On Sat, Mar 16, 2013 at 11:40:19PM +0000, Sami Kerola wrote:
> Now when the whereis will not have search directory duplicates, and it
> finds real path for symlinked files in paths it is time to remove search
> result duplicates.

 Hmm... it seems like unnecessary patch. I guess that

> Addresses: https://bugzilla.redhat.com/show_bug.cgi?id=835211

 is resolved by unique canonicalized directories where we search for
 the files.

   $ ./whereis -b ls
   ls: /usr/bin/ls

 (no more duplicate results like "ls: /bin/ls /usr/bin/ls").

    Karel

-- 
 Karel Zak  <kzak@redhat.com>
 http://karelzak.blogspot.com

Re: [PATCH 0/7] pull: whereis changes

[Karel Zak]

On Sat, Mar 16, 2013 at 11:40:12PM +0000, Sami Kerola wrote:
> Here are the results of the whereis(1) work which I mentioned earlier the
> week in irc.  All of the patches are related to that command, and as you
> can see from the diff stats a lot has changed.

 Almost all applied. After many years I'm able to read whereis.c, thanks!

    Karel


-- 
 Karel Zak  <kzak@redhat.com>
 http://karelzak.blogspot.com

Re: [PATCH 6/7] whereis: canonicalize files before printing

[Sami Kerola]

On Wed, Mar 20, 2013 at 12:56 PM, Karel Zak <kzak@redhat.com> wrote:
> On Sat, Mar 16, 2013 at 11:40:18PM +0000, Sami Kerola wrote:
>> The whereis gives promise to find files so it seems to be more correct
>> print symlink destinations than locations of symlinks.
>
>  Not sure. I think it's correct to canonicalize directory paths (to
>  avoid problems like /bin -> /usr/bin symlinks), but I don't see
>  a reason to canonicalize whole paths (filenames).
>
>  For example
>
>     $ whereis -b bzcat
>     bzcat: /usr/bin/bzcat
>
>  is expected answer, but
>
>     $ whereis -b bzcat
>     bzcat: /usr/bin/bzip2
>
>  looks strange. The another story is that many things are redirected
>  by symlinks to /etc/alternatives and then to pretty specific places.
>
>  Maybe we can add an option to canonicalize, but from my point of view
>  it should not be enabled by default.
>
>  I have applied patch to canonicalize directory names.

Hi Karel,

Thank you for the additional work put to whereis(1) patch series.

I agree. Additional --canonical option is right way to expose where
files are. I will have a look of this sometime later the week(end).

-- 
   Sami Kerola
   http://www.iki.fi/kerolasa/