PATCHES: Device discovery of the "bluetooth" CUPS backend does not work

Hi,

I am using Bluez 4.48 in Ubuntu Karmic. First thank you for applying my patches. The bluetooth backend works perfectly now with CUPS 1.3.x and older. On the new CUPS 1.4.x a new problem appeared: The new CUPS does device discovery only for a given time frame requested by the client (printer setup tool, "lpinfo" command). CUPS's default for CUPS-1.3.x-ish requests without timeout specification seems to be 10 seconds. CUPS startys all backends at once in the beginning (in parallel) and kills every backend which remains running at the end of the timeout. It accepts output from the backends whenever it occurs not only when the backend finishes, so a backend can search for printers infinitely long if it outputs every found device immediately. Then all printers found during CUPS' timeout are taken into account.

The bluetooth backend o 4.48 asks the Bluetooth daemon for printers and collects results for 10 seconds and after that it output them. This takes a total of 10.5 sec and so CUPS kills the backend right before it answers (at least with the 10-second default timeout), resulting in Bluetooth printers never being discovered by CUPS.

I have fixed this now with the attached patch, by making each new printer added to the list being output immediately. Note that the list structure cannot be removed from cups/main.c as otherwise we would get duplicate listings.

Can you please apply this patch to Bluez so that Bluez gets ready for the new CUPS 1.4.x generation? Thanks.

Till

http://www.openprinting.org/

diff -Nur -x '*.orig' -x '*~' bluez-4.48/cups/main.c bluez-4.48.new/cups/main.c
@@ -185,6 +185,27 @@
 	return id;
 }
 
+static void print_printer_details(const char *name, const char *bdaddr, const char *id)
+{
+	char *uri, *escaped;
+
+	escaped = g_strdelimit(g_strdup(name), "\"", '\'');
+	uri = g_strdup_printf("bluetooth://%c%c%c%c%c%c%c%c%c%c%c%c",
+		 bdaddr[0], bdaddr[1],
+		 bdaddr[3], bdaddr[4],
+		 bdaddr[6], bdaddr[7],
+		 bdaddr[9], bdaddr[10],
+		 bdaddr[12], bdaddr[13],
+		 bdaddr[15], bdaddr[16]);
+	printf("direct %s \"%s\" \"%s (Bluetooth)\"", uri, escaped, escaped);
+	if (id != NULL)
+		printf(" \"%s\"\n", id);
+	else
+		printf("\n");
+	g_free(escaped);
+	g_free(uri);
+}
+
 static void add_device_to_list(const char *name, const char *bdaddr, const char *id)
 {
 	struct cups_device *device;

@@ -212,27 +233,7 @@
 	device->id = g_strdup(id);
 
 	device_list = g_slist_prepend(device_list, device);
-}
-
-static void print_printer_details(const char *name, const char *bdaddr, const char *id)
-{
-	char *uri, *escaped;
-
-	escaped = g_strdelimit(g_strdup(name), "\"", '\'');
-	uri = g_strdup_printf("bluetooth://%c%c%c%c%c%c%c%c%c%c%c%c",
-		 bdaddr[0], bdaddr[1],
-		 bdaddr[3], bdaddr[4],
-		 bdaddr[6], bdaddr[7],
-		 bdaddr[9], bdaddr[10],
-		 bdaddr[12], bdaddr[13],
-		 bdaddr[15], bdaddr[16]);
-	printf("direct %s \"%s\" \"%s (Bluetooth)\"", uri, escaped, escaped);
-	if (id != NULL)
-		printf(" \"%s\"\n", id);
-	else
-		printf("\n");
-	g_free(escaped);
-	g_free(uri);
+	print_printer_details(device->name, device->bdaddr, device->id);
 }
 
 static gboolean parse_device_properties(DBusMessageIter *reply_iter, char **name, char **bdaddr)

@@ -384,23 +385,6 @@
 
 static void discovery_completed(void)
 {
-	GSList *l;
-
-	for (l = device_list; l != NULL; l = l->next) {
-		struct cups_device *device = (struct cups_device *) l->data;
-
-		if (device->name == NULL)
-			device->name = g_strdelimit(g_strdup(device->bdaddr), ":", '-');
-		/* Give another try to getting an ID for the device */
-		if (device->id == NULL)
-			remote_device_found(NULL, device->bdaddr, device->name);
-		print_printer_details(device->name, device->bdaddr, device->id);
-		g_free(device->name);
-		g_free(device->bdaddr);
-		g_free(device->id);
-		g_free(device);
-	}
-
 	g_slist_free(device_list);
 	device_list = NULL;
 

@@ -638,6 +622,9 @@
 	/* Make sure status messages are not buffered */
 	setbuf(stderr, NULL);
 
+	/* Make sure output is not buffered */
+	setbuf(stdout, NULL);
+
 	/* Ignore SIGPIPE signals */
 #ifdef HAVE_SIGSET
 	sigset(SIGPIPE, SIG_IGN);

Hi Till,

send a patch created with git format-patch so we can apply them properly without me having to modify the commit message manually.

Regards

Marcel

Thank you for the quick answer.

Patch in GIT format attached.

Till

From: Till Kamppeter <[email protected]>
Date: Wed, 26 Aug 2009 21:05:43 +0200
Subject: [PATCH] Make discovery mode of bluetooth CUPS backend work with CUPS 1.4.0.

---
 ChangeLog   |    2 +
 cups/main.c |   63 +++++++++++++++++++++++-----------------------------------
 2 files changed, 27 insertions(+), 38 deletions(-)

index d06f494..6c170f6 100644
@@ -1,3 +1,5 @@
+ver 4.51:
+	Make discovery mode of bluetooth CUPS backend work with CUPS 1.4.0.
 ver 4.50:
 	Fix issue with missing manual pages in distribution.
 	Fix issue with the configuration and state directories.

1.6.3.3

Hi Till,

please stop top posting. This is a proper open source mailing list and we do proper inline quoting.

So you have time to write a lengthy email with an explanation what is going on, but the commit message is empty. Please fix the commit message since that is where it should be. Especially since you wrote it already anyway.

Regards

Marcel

The new CUPS 1.4.x does device discovery only for a given time frame requested by the client (printer setup tool, "lpinfo" command). CUPS's default for CUPS-1.3.x-ish requests without timeout specification seems to be 10 seconds. CUPS starts all backends at once in the beginning (in parallel) and kills every backend which remains running at the end of the timeout. It accepts output from the backends whenever it occurs not only when the backend finishes, so a backend can search for printers infinitely long if it outputs every found device immediately. Then all printers found during CUPS' timeout are taken into account.

The bluetooth backend of 4.48 asks the Bluetooth daemon for printers and collects results for 10 seconds and after that it outputs them. This takes a total of 10.5 sec and so CUPS kills the backend right before it answers (at least with the 10-second default timeout), resulting in Bluetooth printers never being discovered by CUPS.

This change fixes it by making each new printer added to the list being output immediately. Note that the list structure cannot be removed from cups/main.c as otherwise we would get duplicate listings. Also important is the addition of unbuffered output on stdout.

---
 ChangeLog   |    2 +
 cups/main.c |   63 
+++++++++++++++++++++++-----------------------------------
 2 files changed, 27 insertions(+), 38 deletions(-)

index d06f494..6c170f6 100644
@@ -1,3 +1,5 @@
+ver 4.51:
+	Make discovery mode of bluetooth CUPS backend work with CUPS 1.4.0.
 ver 4.50:
 	Fix issue with missing manual pages in distribution.
 	Fix issue with the configuration and state directories.

index da757b0..1bbc78c 100644
@@ -185,6 +185,27 @@ static char *device_get_ieee1284_id(const char *adapter, const char *device)
 	return id;
 }
 
+static void print_printer_details(const char *name, const char *bdaddr, const char *id)
+{
+	char *uri, *escaped;
+
+	escaped = g_strdelimit(g_strdup(name), "\"", '\'');
+	uri = g_strdup_printf("bluetooth://%c%c%c%c%c%c%c%c%c%c%c%c",
+		 bdaddr[0], bdaddr[1],
+		 bdaddr[3], bdaddr[4],
+		 bdaddr[6], bdaddr[7],
+		 bdaddr[9], bdaddr[10],
+		 bdaddr[12], bdaddr[13],
+		 bdaddr[15], bdaddr[16]);
+	printf("direct %s \"%s\" \"%s (Bluetooth)\"", uri, escaped, escaped);
+	if (id != NULL)
+		printf(" \"%s\"\n", id);
+	else
+		printf("\n");
+	g_free(escaped);
+	g_free(uri);
+}
+
 static void add_device_to_list(const char *name, const char *bdaddr, 

const char *id)
{
struct cups_device *device;
@@ -212,27 +233,7 @@ static void add_device_to_list(const char *name, const char *bdaddr, const char
device->id = g_strdup(id);

device_list = g_slist_prepend(device_list, device);
-}
-
-static void print_printer_details(const char *name, const char *bdaddr, const char *id) -{
- char *uri, *escaped;
-
- escaped = g_strdelimit(g_strdup(name), "\"", '\'');
- uri = g_strdup_printf("bluetooth://%c%c%c%c%c%c%c%c%c%c%c%c",
- bdaddr[0], bdaddr[1],
- bdaddr[3], bdaddr[4],
- bdaddr[6], bdaddr[7],
- bdaddr[9], bdaddr[10],
- bdaddr[12], bdaddr[13],
- bdaddr[15], bdaddr[16]);
- printf("direct %s \"%s\" \"%s (Bluetooth)\"", uri, escaped, escaped);
- if (id != NULL)
- printf(" \"%s\"\n", id);
- else
- printf("\n");
- g_free(escaped);
- g_free(uri);
+ print_printer_details(device->name, device->bdaddr, device->id);
}

static gboolean parse_device_properties(DBusMessageIter *reply_iter,
char **name, char **bdaddr)
@@ -384,23 +385,6 @@ static void remote_device_found(const char *adapter, const char *bdaddr, const c

static void discovery_completed(void)
{
- GSList *l;
-
- for (l = device_list; l != NULL; l = l->next) {
- struct cups_device *device = (struct cups_device *) l->data;
-
- if (device->name == NULL)
- device->name = g_strdelimit(g_strdup(device->bdaddr), ":", '-');
- /* Give another try to getting an ID for the device */
- if (device->id == NULL)
- remote_device_found(NULL, device->bdaddr, device->name);
- print_printer_details(device->name, device->bdaddr, device->id);
- g_free(device->name);
- g_free(device->bdaddr);
- g_free(device->id);
- g_free(device);
- }
-
g_slist_free(device_list);
device_list = NULL;

@@ -638,6 +622,9 @@ int main(int argc, char *argv[])
/* Make sure status messages are not buffered */ setbuf(stderr, NULL);

+ /* Make sure output is not buffered */
+ setbuf(stdout, NULL);
+
/* Ignore SIGPIPE signals */
#ifdef HAVE_SIGSET
sigset(SIGPIPE, SIG_IGN);

1.6.3.3

Hi Till,

I can't use git am to apply this patch. Someone has to fix this for me.

Regards

Marcel

Hi Marcel:

It looks like Till's email client is just beating it up. Applies for me with "git am" and the attached patch fine, but the inline is corrupt.

Mario Limonciello
[email protected]
Sent from Austin, TX, United States

From fdf7b75f5fc8a9726db6756bcddff6aa74b42ba0 Mon Sep 17 00:00:00 2001
From: Till Kamppeter <[email protected]>
Date: Thu, 27 Aug 2009 09:07:49 +0200
Subject: [PATCH] Make discovery mode of bluetooth CUPS backend work with CUPS 1.4.0

The new CUPS 1.4.x does device discovery only for a given time frame requested by the client (printer setup tool, "lpinfo" command). CUPS's default for CUPS-1.3.x-ish requests without timeout specification seems to be 10 seconds. CUPS starts all backends at once in the beginning (in parallel) and kills every backend which remains running at the end of the timeout. It accepts output from the backends whenever it occurs not only when the backend finishes, so a backend can search for printers infinitely long if it outputs every found device immediately. Then all printers found during CUPS' timeout are taken into account.

The bluetooth backend of 4.48 asks the Bluetooth daemon for printers and collects results for 10 seconds and after that it outputs them. This takes a total of 10.5 sec and so CUPS kills the backend right before it answers (at least with the 10-second default timeout), resulting in Bluetooth printers never being discovered by CUPS.

This change fixes it by making each new printer added to the list being output immediately. Note that the list structure cannot be removed from cups/main.c as otherwise we would get duplicate listings. Also important is the addition of unbuffered output on stdout.

---
 ChangeLog   |    2 +
 cups/main.c |   63 +++++++++++++++++++++++-----------------------------------
 2 files changed, 27 insertions(+), 38 deletions(-)

index d06f494..6c170f6 100644
@@ -1,3 +1,5 @@
+ver 4.51:
+	Make discovery mode of bluetooth CUPS backend work with CUPS 1.4.0.
 ver 4.50:
 	Fix issue with missing manual pages in distribution.
 	Fix issue with the configuration and state directories.

1.6.3.3

Hi Marcel:

It looks like Till's email client is just beating it up. Applies for me with "git am" and the attached patch fine, but the inline is corrupt.

--
Mario Limonciello
[email protected]
Sent from Austin, TX, United States

From fdf7b75f5fc8a9726db6756bcddff6aa74b42ba0 Mon Sep 17 00:00:00 2001
From: Till Kamppeter <[email protected]>
Date: Thu, 27 Aug 2009 09:07:49 +0200
Subject: [PATCH] Make discovery mode of bluetooth CUPS backend work with CUPS 1.4.0

The new CUPS 1.4.x does device discovery only for a given time frame requested by the client (printer setup tool, "lpinfo" command). CUPS's default for CUPS-1.3.x-ish requests without timeout specification seems to be 10 seconds. CUPS starts all backends at once in the beginning (in parallel) and kills every backend which remains running at the end of the timeout. It accepts output from the backends whenever it occurs not only when the backend finishes, so a backend can search for printers infinitely long if it outputs every found device immediately. Then all printers found during CUPS' timeout are taken into account.

The bluetooth backend of 4.48 asks the Bluetooth daemon for printers and collects results for 10 seconds and after that it outputs them. This takes a total of 10.5 sec and so CUPS kills the backend right before it answers (at least with the 10-second default timeout), resulting in Bluetooth printers never being discovered by CUPS.

This change fixes it by making each new printer added to the list being output immediately. Note that the list structure cannot be removed from cups/main.c as otherwise we would get duplicate listings. Also important is the addition of unbuffered output on stdout.

---
 ChangeLog   |    2 +
 cups/main.c |   63 +++++++++++++++++++++++-----------------------------------
 2 files changed, 27 insertions(+), 38 deletions(-)

index d06f494..6c170f6 100644
@@ -1,3 +1,5 @@
+ver 4.51:
+	Make discovery mode of bluetooth CUPS backend work with CUPS 1.4.0.
 ver 4.50:
 	Fix issue with missing manual pages in distribution.
 	Fix issue with the configuration and state directories.

1.6.3.3

Hi Mario,

can I get this one without touching ChangeLog. That file will only be touched be me before making the release. You guys make it pretty hard for me to apply patches. It is not that complicated.

Regards

Marcel

Looks fine to me, although I'd have the stdout unbuffered change in a separate patch instead.

Cheers