Re: [Dovecot] Patch for pigeonhole 0.4.0 avoiding PATH_MAX

[Svante Signell]

On Tue, 2013-06-18 at 12:25 +0200, Svante Signell wrote:
> On Tue, 2013-06-18 at 08:37 +0200, Stephan Bosch wrote:
> > On 6/15/2013 3:24 PM, Svante Signell wrote:
> > > Hi,
> > >
> > > I recently downloaded and built dovecot-2.2.2 and
> > > dovecot-2.2-pigeonhole-0.4.0 on GNU/Linux and GNU/Hurd. The changes
> > > needed will be sent to the Debian maintainer shortly. Latest Debian
> > > release is 2.1.7-7 and dovecot-2.1-pigeonhole-0.3.1. When building
> > > dovecot-2.2.2 there were no PATH_MAX problems on GNU/Hurd, thank you for
> > > that. However, pigeonhole 0.4.0 had one remaining PATH_MAX construct.
> > > The attached patch solves this problem. It it good enough to be accepted
> > > upstream? (According to the description of t_malloc, free is not needed,
> > > right?)
> > 
> > Fixed:
> > 
> > http://hg.rename-it.nl/dovecot-2.2-pigeonhole/rev/1b1a0c271383
> 
> Thanks a lot :)

Are you planning to make a new (minor) release of pigeonhole soon, just
to know what to submit in the bug report to the Debian maintainer?

Re: [Dovecot] Patch for pigeonhole 0.4.0 avoiding PATH_MAX

[Svante Signell]

On Tue, 2013-06-18 at 08:37 +0200, Stephan Bosch wrote:
> On 6/15/2013 3:24 PM, Svante Signell wrote:
> > Hi,
> >
> > I recently downloaded and built dovecot-2.2.2 and
> > dovecot-2.2-pigeonhole-0.4.0 on GNU/Linux and GNU/Hurd. The changes
> > needed will be sent to the Debian maintainer shortly. Latest Debian
> > release is 2.1.7-7 and dovecot-2.1-pigeonhole-0.3.1. When building
> > dovecot-2.2.2 there were no PATH_MAX problems on GNU/Hurd, thank you for
> > that. However, pigeonhole 0.4.0 had one remaining PATH_MAX construct.
> > The attached patch solves this problem. It it good enough to be accepted
> > upstream? (According to the description of t_malloc, free is not needed,
> > right?)
> 
> Fixed:
> 
> http://hg.rename-it.nl/dovecot-2.2-pigeonhole/rev/1b1a0c271383

Thanks a lot :)

[Dovecot] Patch for pigeonhole 0.4.0 avoiding PATH_MAX

[Svante Signell]

Hi,

I recently downloaded and built dovecot-2.2.2 and
dovecot-2.2-pigeonhole-0.4.0 on GNU/Linux and GNU/Hurd. The changes
needed will be sent to the Debian maintainer shortly. Latest Debian
release is 2.1.7-7 and dovecot-2.1-pigeonhole-0.3.1. When building
dovecot-2.2.2 there were no PATH_MAX problems on GNU/Hurd, thank you for
that. However, pigeonhole 0.4.0 had one remaining PATH_MAX construct.
The attached patch solves this problem. It it good enough to be accepted
upstream? (According to the description of t_malloc, free is not needed,
right?)

Thanks,
Svante Signell

Index: dovecot-2.2.2/pigeonhole/src/lib-sievestorage/sieve-storage-script.c
===
--- dovecot-2.2.2.orig/pigeonhole/src/lib-sievestorage/sieve-storage-script.c	2013-06-15 15:12:58.0 +0200
+++ dovecot-2.2.2/pigeonhole/src/lib-sievestorage/sieve-storage-script.c	2013-06-15 15:18:34.0 +0200
@@ -118,12 +118,24 @@
 static int sieve_storage_read_active_link
 (struct sieve_storage *storage, const char **link_r)
 {
-  char linkbuf[PATH_MAX];
+  struct stat st;
+  char *linkbuf;
   int ret;
 
 	*link_r = NULL;
 
-	ret = readlink(storage->active_path, linkbuf, sizeof(linkbuf));
+	/* Stat the file */
+	if ( lstat(storage->active_path, &st) != 0 ) {
+		if ( errno != ENOENT ) {
+			sieve_storage_set_critical(storage,
+"Failed to stat active sieve script symlink (%s): %m.",
+storage->active_path);
+			return -1;
+		}
+	}
+	linkbuf = t_malloc(st.st_size + 1);
+
+	ret = readlink(storage->active_path, linkbuf, st.st_size+1);
 
 	if ( ret < 0 ) {
 		if ( errno == EINVAL ) {