[Printing-architecture] [patch] CUPS 1.7.0: fix corrupted output for USB printers

[Printing-architecture] [patch] CUPS 1.7.0: fix corrupted output for USB printers

[Tim Waugh]

[-- Attachment #1.1: Type: text/plain, Size: 809 bytes --]

When using the usb backend, if a usb-1 printer connected to a usb-2 hub
runs out of paper during the job the print output will be corrupted.
https://bugzilla.redhat.com/show_bug.cgi?id=1026914

This is due to using a 60 second timeout for the libusb_bulk_transfer()
call. Timeout handling and retries are not reliable in every case, as
described here:
https://bugzilla.redhat.com/show_bug.cgi?id=1035863

Quoting from that report:
==>
So TL;DR: this a known and unfixable problem, and the solution is to
simply not cancel transfers (by setting a timeout) and then decide you
want to finish the transfer after all. Either decide you want to really
cancel the transfer, or just wait for it to finish normally.
<==

Patch attached to change the timeout to 0 to wait indefinitely.

Tim.
*/


[-- Attachment #1.2: cups-usb-timeout.patch --]
[-- Type: text/x-patch, Size: 1107 bytes --]

diff -up cups-1.7.0/backend/usb-libusb.c.usb-timeout cups-1.7.0/backend/usb-libusb.c
--- cups-1.7.0/backend/usb-libusb.c.usb-timeout	2013-07-17 16:51:43.000000000 +0100
+++ cups-1.7.0/backend/usb-libusb.c	2013-11-28 16:44:03.785124015 +0000
@@ -496,7 +496,7 @@ print_device(const char *uri,		/* I - De
 	iostatus = libusb_bulk_transfer(g.printer->handle,
 					g.printer->write_endp,
 					print_buffer, g.print_bytes,
-					&bytes, 60000);
+					&bytes, 0);
        /*
 	* Ignore timeout errors, but retain the number of bytes written to
 	* avoid sending duplicate data...
@@ -519,7 +519,7 @@ print_device(const char *uri,		/* I - De
 	  iostatus = libusb_bulk_transfer(g.printer->handle,
 					  g.printer->write_endp,
 					  print_buffer, g.print_bytes,
-					  &bytes, 60000);
+					  &bytes, 0);
 	}
 
        /*
@@ -534,7 +534,7 @@ print_device(const char *uri,		/* I - De
 	  iostatus = libusb_bulk_transfer(g.printer->handle,
 					  g.printer->write_endp,
 					  print_buffer, g.print_bytes,
-					  &bytes, 60000);
+					  &bytes, 0);
         }
 
 	if (iostatus)

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 482 bytes --]

Re: [Printing-architecture] [patch] CUPS 1.7.0: fix corrupted output for USB printers

[Michael Sweet]

Tim,

On Nov 29, 2013, at 7:08 AM, Tim Waugh <twaugh@redhat.com> wrote:
> When using the usb backend, if a usb-1 printer connected to a usb-2 hub
> runs out of paper during the job the print output will be corrupted.
> https://bugzilla.redhat.com/show_bug.cgi?id=1026914
> 
> This is due to using a 60 second timeout for the libusb_bulk_transfer()
> call. Timeout handling and retries are not reliable in every case, as
> described here:
> https://bugzilla.redhat.com/show_bug.cgi?id=1035863
> 
> Quoting from that report:
> ==>
> So TL;DR: this a known and unfixable problem, and the solution is to
> simply not cancel transfers (by setting a timeout) and then decide you
> want to finish the transfer after all. Either decide you want to really
> cancel the transfer, or just wait for it to finish normally.
> <==
> 
> Patch attached to change the timeout to 0 to wait indefinitely.

This mirrors what we do on OS X; will include this in CUPS 1.7.1.

Filed as:

    <rdar://problem/15564888> cups.org: Fix USB timeouts on Linux

_______________________________________________________________
Michael Sweet, Senior Printing System Engineer, PWG Chair