Skip to contentSkip to conversation

exterminate strtok - usr/gen_init_cpio.c

PATCH In review Jesper Juhl 6 participants 11 messages Aug 24, 2005 – Aug 26, 2005 Kbuild

lore
Share & tools
lore
#0 lore raw

PATCH 3/3 exterminate strtok - usr/gen_init_cpio.c

Convert strtok() use to strsep() in usr/gen_init_cpio.c

I've compile tested this patch and it compiles fine.
I build a 2.6.13-rc6-mm2 kernel with the patch applied without problems, and the resulting kernel boots and runs just fine (using it right now). But despite this basic testing it would still be nice if someone would double-check that I haven't made some silly mistake that would break some other setup than mine.

---

 gen_init_cpio.c |   31 ++++++++++++++++++++++---------
 1 files changed, 22 insertions(+), 9 deletions(-)
usr/gen_init_cpio.c +22 -9
@@ -438,7 +438,7 @@ struct file_handler file_handler_table[]
 int main (int argc, char *argv[])
 {
 	FILE *cpio_list;
-	char line[LINE_SIZE];
+	char *line, *ln;
 	char *args, *type;
 	int ec = 0;
 	int line_nr = 0;
@@ -455,7 +455,14 @@ int main (int argc, char *argv[])
 		exit(1);
 	}
 
-	while (fgets(line, LINE_SIZE, cpio_list)) {
+	ln = malloc(LINE_SIZE);
+	if (!ln) {
+		fprintf(stderr,
+			"ERROR: unable to allocate %d bytes of buffer\n", LINE_SIZE);
+		exit(1);
+	}
+
+	while (line = ln, fgets(line, LINE_SIZE, cpio_list)) {
 		int type_idx;
 		size_t slen = strlen(line);
 
@@ -466,14 +473,15 @@ int main (int argc, char *argv[])
 			continue;
 		}
 
-		if (! (type = strtok(line, " \t"))) {
+		if (! (type = strsep(&line, " \t"))) {
 			fprintf(stderr,
 				"ERROR: incorrect format, could not locate file type line %d: '%s'\n",
-				line_nr, line);
+				line_nr, ln);
 			ec = -1;
+			continue;
 		}
 
-		if ('\n' == *type) {
+		if (!*type || '\n' == *type) {
 			/* a blank line */
 			continue;
 		}
@@ -483,16 +491,20 @@ int main (int argc, char *argv[])
 			continue;
 		}
 
-		if (! (args = strtok(NULL, "\n"))) {
+		if (! (args = strsep(&line, "\n"))) {
 			fprintf(stderr,
 				"ERROR: incorrect format, newline required line %d: '%s'\n",
-				line_nr, line);
+				line_nr, ln);
 			ec = -1;
+			continue;
 		}
+		
+		if (!*args)
+			continue;
 
 		for (type_idx = 0; file_handler_table[type_idx].type; type_idx++) {
 			int rc;
-			if (! strcmp(line, file_handler_table[type_idx].type)) {
+			if (! strcmp(type, file_handler_table[type_idx].type)) {
 				if ((rc = file_handler_table[type_idx].handler(args))) {
 					ec = rc;
 					fprintf(stderr, " line %d\n", line_nr);
@@ -503,9 +515,10 @@ int main (int argc, char *argv[])
 
 		if (NULL == file_handler_table[type_idx].type) {
 			fprintf(stderr, "unknown file type line %d: '%s'\n",
-				line_nr, line);
+				line_nr, ln);
 		}
 	}
+	free(ln);
 	cpio_trailer();
 
 	exit(ec);
#1 lore raw
Quoted 26 lines · Jesper Juhl

Convert strtok() use to strsep() in usr/gen_init_cpio.c

I've compile tested this patch and it compiles fine.
I build a 2.6.13-rc6-mm2 kernel with the patch applied without problems, and
the resulting kernel boots and runs just fine (using it right now).
But despite this basic testing it would still be nice if someone would double-check that I haven't made some silly mistake that would break some other setup than mine.

---

 gen_init_cpio.c |   31 ++++++++++++++++++++++---------
 1 files changed, 22 insertions(+), 9 deletions(-)
usr/gen_init_cpio.c +1 -1
@@ -438,7 +438,7 @@ struct file_handler file_handler_table[]
 int main (int argc, char *argv[])
 {
 	FILE *cpio_list;
-	char line[LINE_SIZE];
+	char *line, *ln;
 	char *args, *type;
 	int ec = 0;
 	int line_nr = 0;

usr/gen_init_cpio.c int main (int argc, char *argv[]) · line 458

Quoted 5 lines
usr/gen_init_cpio.c +1 -1
@@ -455,7 +455,14 @@ int main (int argc, char *argv[])
 		exit(1);
 	}
 
-	while (fgets(line, LINE_SIZE, cpio_list)) {
+	ln = malloc(LINE_SIZE);

Why change to malloc()? This is a userspace program. It doesn't have the kernel stack constraints.

--
Brian Gerst

#2 lore raw
Quoted 33 lines · Brian Gerst
Quoted 31 lines · Jesper Juhl

Convert strtok() use to strsep() in usr/gen_init_cpio.c

I've compile tested this patch and it compiles fine.
I build a 2.6.13-rc6-mm2 kernel with the patch applied without problems, and
the resulting kernel boots and runs just fine (using it right now).
But despite this basic testing it would still be nice if someone would double-check that I haven't made some silly mistake that would break some other setup than mine.

---

 gen_init_cpio.c |   31 ++++++++++++++++++++++---------
 1 files changed, 22 insertions(+), 9 deletions(-)

--- linux-2.6.13-rc6-mm2-orig/usr/gen_init_cpio.c 2005-06-17
21:48:29.000000000 +0200
+++ linux-2.6.13-rc6-mm2/usr/gen_init_cpio.c 2005-08-24
18:58:21.000000000 +0200

usr/gen_init_cpio.c +1 -1
@@ -438,7 +438,7 @@ struct file_handler file_handler_table[]
 int main (int argc, char *argv[])
 {
     FILE *cpio_list;
-    char line[LINE_SIZE];
+    char *line, *ln;
     char *args, *type;
     int ec = 0;
     int line_nr = 0;
@@ -455,7 +455,14 @@ int main (int argc, char *argv[])
         exit(1);
     }
 
Quoted 5 lines

usr/gen_init_cpio.c int main (int argc, char *argv[]) · line 458

Quoted 2 lines
usr/gen_init_cpio.c +1 -1
-    while (fgets(line, LINE_SIZE, cpio_list)) {
+    ln = malloc(LINE_SIZE);

Why change to malloc()? This is a userspace program. It doesn't have the kernel stack constraints.

Good catch, agreed.

I prefer the code as-is, with LINE_SIZE stack allocations.

Jeff

#3 lore raw
Quoted 44 lines · Jeff Garzik
Quoted 36 lines · Brian Gerst
Quoted 34 lines · Jesper Juhl

Convert strtok() use to strsep() in usr/gen_init_cpio.c

I've compile tested this patch and it compiles fine.
I build a 2.6.13-rc6-mm2 kernel with the patch applied without problems, and
the resulting kernel boots and runs just fine (using it right now). But despite this basic testing it would still be nice if someone would double-check that I haven't made some silly mistake that would break some other setup than mine.

---

 gen_init_cpio.c |   31 ++++++++++++++++++++++---------
 1 files changed, 22 insertions(+), 9 deletions(-)

--- linux-2.6.13-rc6-mm2-orig/usr/gen_init_cpio.c 2005-06-17
21:48:29.000000000 +0200
+++ linux-2.6.13-rc6-mm2/usr/gen_init_cpio.c 2005-08-24
18:58:21.000000000 +0200

usr/gen_init_cpio.c +1 -1
@@ -438,7 +438,7 @@ struct file_handler file_handler_table[]
 int main (int argc, char *argv[])
 {
     FILE *cpio_list;
-    char line[LINE_SIZE];
+    char *line, *ln;
     char *args, *type;
     int ec = 0;
     int line_nr = 0;
@@ -455,7 +455,14 @@ int main (int argc, char *argv[])
         exit(1);
     }
 
Quoted 6 lines

usr/gen_init_cpio.c int main (int argc, char *argv[]) · line 458

Quoted 2 lines
usr/gen_init_cpio.c +1 -1
-    while (fgets(line, LINE_SIZE, cpio_list)) {
+    ln = malloc(LINE_SIZE);

Why change to malloc()? This is a userspace program. It doesn't have the kernel stack constraints.

Quoted 4 lines

Good catch, agreed.

I prefer the code as-is, with LINE_SIZE stack allocations.

The reason I did it like that was that strsep takes offense at strsep(&line, ...) when line is allocated on the stack. So I just changed it around to being malloc()'ed and things were good.

Signature · 3 lines
Jesper Juhl <[email protected]>
Don't top-post  http://www.catb.org/~esr/jargon/html/T/top-post.html
Plain text mails only, please      http://www.expita.com/nomime.html
#4 lore raw
Quoted 48 lines · Jesper Juhl
Quoted 46 lines · Jeff Garzik
Quoted 43 lines · Brian Gerst
Quoted 31 lines · Jesper Juhl

Convert strtok() use to strsep() in usr/gen_init_cpio.c

I've compile tested this patch and it compiles fine.
I build a 2.6.13-rc6-mm2 kernel with the patch applied without problems, and
the resulting kernel boots and runs just fine (using it right now). But despite this basic testing it would still be nice if someone would double-check that I haven't made some silly mistake that would break some other setup than mine.

---

gen_init_cpio.c | 31 ++++++++++++++++++++++---------

1 files changed, 22 insertions(+), 9 deletions(-)

--- linux-2.6.13-rc6-mm2-orig/usr/gen_init_cpio.c 2005-06-17
21:48:29.000000000 +0200
+++ linux-2.6.13-rc6-mm2/usr/gen_init_cpio.c 2005-08-24
18:58:21.000000000 +0200

usr/gen_init_cpio.c +1 -1
@@ -438,7 +438,7 @@ struct file_handler file_handler_table[]
 int main (int argc, char *argv[])
 {
    FILE *cpio_list;
-    char line[LINE_SIZE];
+    char *line, *ln;
    char *args, *type;
    int ec = 0;
    int line_nr = 0;

usr/gen_init_cpio.c int main (int argc, char *argv[]) · line 458

Quoted 5 lines
usr/gen_init_cpio.c +1 -1
@@ -455,7 +455,14 @@ int main (int argc, char *argv[])
        exit(1);
    }
 
-    while (fgets(line, LINE_SIZE, cpio_list)) {
+    ln = malloc(LINE_SIZE);

Why change to malloc()? This is a userspace program. It doesn't have the kernel stack constraints.

Quoted 9 lines
Quoted 4 lines

Good catch, agreed.

I prefer the code as-is, with LINE_SIZE stack allocations.

The reason I did it like that was that strsep takes offense at strsep(&line, ...) when line is allocated on the stack. So I just changed it around to being malloc()'ed and things were good.

Do this instead:

	char ln[LINE_SIZE], *line;

--
Brian Gerst

#5 lore raw
Quoted 4 lines · Brian Gerst

Do this instead:

	char ln[LINE_SIZE], *line;

Right, now why didn't I think of that :)

Jeff: Does the patch below agree with you more?

---

 usr/gen_init_cpio.c |   22 ++++++++++++++--------
 1 files changed, 14 insertions(+), 8 deletions(-)

--- linux-2.6.13-rc6-mm2-orig/usr/gen_init_cpio.c 2005-06-17 21:48:29.000000000 +0200
+++ linux-2.6.13-rc6-mm2/usr/gen_init_cpio.c 2005-08-24 23:10:36.000000000 +0200
@@ -438,7 +438,8 @@ struct file_handler file_handler_table[]
int main (int argc, char *argv[])
{
FILE *cpio_list;
- char line[LINE_SIZE];
+ char ln[LINE_SIZE];
+ char *line;
char *args, *type;
int ec = 0;
int line_nr = 0;
@@ -455,7 +456,7 @@ int main (int argc, char *argv[])
exit(1);
}

- while (fgets(line, LINE_SIZE, cpio_list)) {
+ while (line = ln, fgets(line, LINE_SIZE, cpio_list)) {
int type_idx;
size_t slen = strlen(line);

@@ -466,11 +467,12 @@ int main (int argc, char *argv[])
continue;
}

- if (! (type = strtok(line, " \t"))) {
+ if (! (type = strsep(&line, " \t"))) {
fprintf(stderr,
"ERROR: incorrect format, could not locate file type line %d: '%s'\n",
- line_nr, line);
+ line_nr, ln);
ec = -1;
+ continue;
}

if ('\n' == *type) {
@@ -483,16 +485,20 @@ int main (int argc, char *argv[])
continue;
}

- if (! (args = strtok(NULL, "\n"))) {
+ if (! (args = strsep(&line, "\n"))) {
fprintf(stderr,
"ERROR: incorrect format, newline required line %d: '%s'\n",
- line_nr, line);
+ line_nr, ln);
ec = -1;
+ continue;
}
+
+ if (!*args)
+ continue;

for (type_idx = 0; file_handler_table[type_idx].type; type_idx++) {
int rc;
- if (! strcmp(line, file_handler_table[type_idx].type)) {
+ if (! strcmp(type, file_handler_table[type_idx].type)) {
if ((rc = file_handler_table[type_idx].handler(args))) {
ec = rc;
fprintf(stderr, " line %d\n", line_nr);
@@ -503,7 +509,7 @@ int main (int argc, char *argv[])

if (NULL == file_handler_table[type_idx].type) {
fprintf(stderr, "unknown file type line %d: '%s'\n",
- line_nr, line);
+ line_nr, ln);
}
}
cpio_trailer();

#6 lore raw
Quoted 29 lines · Jesper Juhl
Quoted 4 lines · Brian Gerst

Do this instead:

	char ln[LINE_SIZE], *line;

Right, now why didn't I think of that :)

Jeff: Does the patch below agree with you more?

---

 usr/gen_init_cpio.c |   22 ++++++++++++++--------
 1 files changed, 14 insertions(+), 8 deletions(-)

--- linux-2.6.13-rc6-mm2-orig/usr/gen_init_cpio.c 2005-06-17 21:48:29.000000000 +0200
+++ linux-2.6.13-rc6-mm2/usr/gen_init_cpio.c 2005-08-24 23:10:36.000000000 +0200

usr/gen_init_cpio.c +2 -1
@@ -438,7 +438,8 @@ struct file_handler file_handler_table[]
 int main (int argc, char *argv[])
 {
 	FILE *cpio_list;
-	char line[LINE_SIZE];
+	char ln[LINE_SIZE];
+	char *line;
 	char *args, *type;
 	int ec = 0;
 	int line_nr = 0;

usr/gen_init_cpio.c int main (int argc, char *argv[]) · line 461

Quoted 7 lines
usr/gen_init_cpio.c +1 -1
@@ -455,7 +456,7 @@ int main (int argc, char *argv[])
 		exit(1);
 	}
 
-	while (fgets(line, LINE_SIZE, cpio_list)) {
+	while (line = ln, fgets(line, LINE_SIZE, cpio_list)) {
 		int type_idx;
 		size_t slen = strlen(line);

The ',' in the while() isn't exactly readable... first order of bussiness inside:

	while (fgets(line, LINE_SIZE, cpio_list)) {
	while (fgets(ln, LINE_SIZE, cpio_list)) {
		int type_idx;
		size_t slen = strlen(ln);

		line = ln;

Or just use the fact that fgets(3) returns the buffer if no error:

	while(line = fgets(ln, LINE_SIZE, cpio_list)) {

(yes, gcc will complain... wrap in () to shut it up)

Signature · 4 lines
Dr. Horst H. von Brand                   User #22616 counter.li.org
Departamento de Informatica                     Fono: +56 32 654431
Universidad Tecnica Federico Santa Maria              +56 32 654239
Casilla 110-V, Valparaiso, Chile                Fax:  +56 32 797513
#7 lore raw
In reply to Jesper Juhl
Quoted 1 line · Jesper Juhl

Convert strtok() use to strsep() in usr/gen_init_cpio.c

This is userland code...

No, I'm not looking it over carfully, just a fast look over.

Quoted 1 line

I've compile tested this patch and it compiles fine.

You should be able ti test it then.

Quoted 16 lines

I build a 2.6.13-rc6-mm2 kernel with the patch applied without problems, and the resulting kernel boots and runs just fine (using it right now). But despite this basic testing it would still be nice if someone would double-check that I haven't made some silly mistake that would break some other setup than mine.

---

 gen_init_cpio.c |   31 ++++++++++++++++++++++---------
 1 files changed, 22 insertions(+), 9 deletions(-)

usr/gen_init_cpio.c struct file_handler file_handler_table[] · line 441

Quoted 5 lines
usr/gen_init_cpio.c +1 -1
@@ -438,7 +438,7 @@ struct file_handler file_handler_table[]
 int main (int argc, char *argv[])
 {
 	FILE *cpio_list;
-	char line[LINE_SIZE];
+	char *line, *ln;

No need for this, there is no particular stack frugality requirement with user code.

usr/gen_init_cpio.c int main (int argc, char *argv[]) · line 458

Quoted 9 lines
usr/gen_init_cpio.c +1 -1
 	char *args, *type;
 	int ec = 0;
 	int line_nr = 0;
@@ -455,7 +455,14 @@ int main (int argc, char *argv[])
 		exit(1);
 	}
 
-	while (fgets(line, LINE_SIZE, cpio_list)) {
+	ln = malloc(LINE_SIZE);

Ditto.

[...]

usr/gen_init_cpio.c int main (int argc, char *argv[]) · line 484

Quoted 2 lines
usr/gen_init_cpio.c +1 -1
-		if ('\n' == *type) {
+		if (!*type || '\n' == *type) {

Redundant. If *type == '\n', it is certainly != 0.

Signature · 4 lines
Dr. Horst H. von Brand                   User #22616 counter.li.org
Departamento de Informatica                     Fono: +56 32 654431
Universidad Tecnica Federico Santa Maria              +56 32 654239
Casilla 110-V, Valparaiso, Chile                Fax:  +56 32 797513
#8 lore raw
Quoted 1 line · Horst von Brand

Jesper Juhl <[email protected]> wrote:

[snip]

Quoted 5 lines

usr/gen_init_cpio.c int main (int argc, char *argv[]) · line 484

Quoted 3 lines
usr/gen_init_cpio.c +1 -1
-             if ('\n' == *type) {
+             if (!*type || '\n' == *type) {

Redundant. If *type == '\n', it is certainly != 0.

Hmm, I added that since if we get past the if above it, then "type" might be !=NULL, but *type might be, but you are right, it's still redundant since there's a check below that'll catch it if so.

Signature · 3 lines
Jesper Juhl <[email protected]>
Don't top-post  http://www.catb.org/~esr/jargon/html/T/top-post.html
Plain text mails only, please      http://www.expita.com/nomime.html
#9 lore raw
In reply to Horst von Brand
Quoted 4 lines · Horst

usr/gen_init_cpio.c int main (int argc, char *argv[]) · line 484

Quoted 2 lines
usr/gen_init_cpio.c +1 -1
-		if ('\n' == *type) {
+		if (!*type || '\n' == *type) {

Redundant. If *type == '\n', it is certainly != 0.

No - I don't think redundant, at least not this change in isolation. Perhaps redundant in light of subsequent code lines, as Jesper notes in his followup.

But it is confusing to read - poor and inconsistent choice of code phrasing.

If the patch had read as:

    -		if (*type == '\n') {
    +		if (*type == '\n' || *type == '\0') {

then it would be clearer to the reader in my view. A check for newline is changed to a check for newline or nul-byte.

(Yes - I recognize that one is not given the freedom to change the -old- lines in a patch for the sake of clarity ;).

Signature · 3 lines
                  I won't rest till it's the best ...
                  Programmer, Linux Scalability
                  Paul Jackson <[email protected]> 1.925.600.0401
#10 lore raw
In reply to Jesper Juhl
Quoted 8 lines · Jesper Juhl

Convert strtok() use to strsep() in usr/gen_init_cpio.c

I've compile tested this patch and it compiles fine.
I build a 2.6.13-rc6-mm2 kernel with the patch applied without problems, and the resulting kernel boots and runs just fine (using it right now). But despite this basic testing it would still be nice if someone would double-check that I haven't made some silly mistake that would break some other setup than mine.

This is userland - I see no point in uglyfying the code.

Sam