threads / patch / 479

patchFix memory leaks in read_tree_recursive()

Subject: [PATCH] Fix memory leaks in read_tree_recursive()

## tl;dr

4 messages between May 4, 2005 and May 5, 2005. Diffs are folded; open one to read it.

replies: 3people: 2as markdown or json

Jonas Fonseca· May 4, 2005, 23:19 UTC · lore

This patch fixes memory leaks in the error path of read_tree_recursive().

Signed-off-by: Jonas Fonseca <fonseca@diku.dk>

--- 571c0f4599fbeefd995bbc24480add1575c36c94/tree.c (mode:100644 sha1:4a26603f6e32866c0db8a01ac1c228be801f76c6) +++ uncommitted/tree.c (mode:100644)

Show changes to diff +6 −2

@@ -39,14 +39,18 @@
 		if (S_ISDIR(mode)) {
 			int retval;
 			int pathlen = strlen(path);
-			char *newbase = xmalloc(baselen + 1 + pathlen);
+			char *newbase;
 			void *eltbuf;
 			char elttype[20];
 			unsigned long eltsize;
 
 			eltbuf = read_sha1_file(sha1, elttype, &eltsize);
-			if (!eltbuf || strcmp(elttype, "tree"))
+			if (!eltbuf || strcmp(elttype, "tree")) {
+				if (eltbuf) mem_free(eltbuf);
 				return -1;
+			}
+
+			newbase = xmalloc(baselen + 1 + pathlen);
 			memcpy(newbase, base, baselen);
 			memcpy(newbase + baselen, path, pathlen);
 			newbase[baselen + pathlen] = '/';

-- 
Jonas Fonseca

Jonas Fonseca· May 4, 2005, 23:39 UTC · re: Jonas Fonseca · lore

This patch fixes memory leaks in read_tree_recursive().

Signed-off-by: Jonas Fonseca <fonseca@diku.dk>

---

Sorry, typo in the previous patch. I will try to be more careful. This patch has already been applied to cogito.

--- 571c0f4599fbeefd995bbc24480add1575c36c94/tree.c (mode:100644 sha1:4a26603f6e32866c0db8a01ac1c228be801f76c6) +++ uncommitted/tree.c (mode:100644)

Show changes to diff +6 −2

@@ -39,14 +39,18 @@
 		if (S_ISDIR(mode)) {
 			int retval;
 			int pathlen = strlen(path);
-			char *newbase = xmalloc(baselen + 1 + pathlen);
+			char *newbase;
 			void *eltbuf;
 			char elttype[20];
 			unsigned long eltsize;
 
 			eltbuf = read_sha1_file(sha1, elttype, &eltsize);
-			if (!eltbuf || strcmp(elttype, "tree"))
+			if (!eltbuf || strcmp(elttype, "tree")) {
+				if (eltbuf) free(eltbuf);
 				return -1;
+			}
+
+			newbase = xmalloc(baselen + 1 + pathlen);
 			memcpy(newbase, base, baselen);
 			memcpy(newbase + baselen, path, pathlen);
 			newbase[baselen + pathlen] = '/';

-- 
Jonas Fonseca

Junio C Hamano· May 4, 2005, 23:43 UTC · re: Jonas Fonseca · lore

Re: [PATCH] Fix memory leaks in read_tree_recursive()

>>>>> "JF" == Jonas Fonseca <fonseca@diku.dk> writes:

JF> This patch fixes memory leaks in the error path of JF> read_tree_recursive().

The leak seems to be real but what is "mem_free"? Has it been compile tested?

JF> @@ -39,14 +39,18 @@
JF>  		if (S_ISDIR(mode)) {
JF>  			int retval;
 
JF> ...
JF> -			if (!eltbuf || strcmp(elttype, "tree"))
JF> +			if (!eltbuf || strcmp(elttype, "tree")) {
JF> +				if (eltbuf) mem_free(eltbuf);
JF>  				return -1;

Btw, who is putting this header in your mail? It does not make sense to me unless Jonas is pseudonym for Linus...

  Mail-Followup-To: Linus Torvalds <torvalds@osdl.org>, git@vger.kernel.org

Jonas Fonseca· May 5, 2005, 00:08 UTC · re: Junio C Hamano · lore

Re: [PATCH] Fix memory leaks in read_tree_recursive()

Junio C Hamano <junkio@cox.net> wrote Wed, May 04, 2005:

Show 6 quoted lines

> >>>>> "JF" == Jonas Fonseca <fonseca@diku.dk> writes:
> 
> JF> This patch fixes memory leaks in the error path of
> JF> read_tree_recursive().
> 
> The leak seems to be real but what is "mem_free"?

That mem_free() was a bad habbit from the project I usually work on.

> Has it been compile tested?

No, but the second patch has.

BTW, when compiling git on a FreeBSD box I get these warnings:

date.c: In function `parse_date': date.c:414: warning: long unsigned int format, time_t arg (arg 4) date.c:414: warning: long unsigned int format, time_t arg (arg 4) date.c: In function `datestamp': date.c:427: warning: long unsigned int format, time_t arg (arg 4) date.c:427: warning: long unsigned int format, time_t arg (arg 4) tar-tree.c: In function `write_header': tar-tree.c:249: warning: long unsigned int format, time_t arg (arg 3) local-pull.c: In function `fetch': local-pull.c:73: warning: long int format, different type arg (arg 4)

because time_t is defined as int32_t. Don't know if they are worth fixing at this point.

Show 14 quoted lines

> JF> @@ -39,14 +39,18 @@
> JF>  		if (S_ISDIR(mode)) {
> JF>  			int retval;
>  
> JF> ...
> JF> -			if (!eltbuf || strcmp(elttype, "tree"))
> JF> +			if (!eltbuf || strcmp(elttype, "tree")) {
> JF> +				if (eltbuf) mem_free(eltbuf);
> JF>  				return -1;
> 
> Btw, who is putting this header in your mail?  It does not make
> sense to me unless Jonas is pseudonym for Linus...
> 
>   Mail-Followup-To: Linus Torvalds <torvalds@osdl.org>, git@vger.kernel.org

Maybe because Mutt knows I am subscribe to this mailing list and assumes I don't want to have mail addressed directly to my email address?

I just hit 'g'.

-- 
Jonas Fonseca