Bug 58925

i can confirm that the fix is working. i was having probelms using a HP 
Photosmart camera with the PTP driver, was receiving timeouts...
been searching for a fix for 2 weeks, when Mariusz Woloszyn told me that 
they found this bug... after editing by hand the libusb sources, and 
recompiling, everything worked ok, without any other software update or 
smtg else...

i did not notice any other problems since than, other than little 
synchronization problems when stressing the camera/USB by turning it 
off, then back on quickly and running gphoto2 --auto-detect in the same 
time on a 400MHz (old!) machine... but i think this is not at all 
related to the patch, which is working fine. i'd recommend patching the 
current port of libusb so that this problem is fixed and posting a note 
or smtg about this, as i've seen at least a dozen people having posted 
related problems in various forums, from Germany to Australia... maybe a 
notice on the gPhoto homepage would be a good idea, if there's any need 
i can write a short description/how-to on this thema, publish it on my 
page and somebody can put a link to it there...

anyway, thanks to all the people working on this!

and if i should write the how-to (only if there's interest in linking 
from the gPhoto2-page to it), just mail me and let me know... if not, 
it's futile work in my opinion, and i shall not do it... i could write a 
fairly complete tutorial on using ur digital camera through the PTP 
driver on FreeBSD, with all the probs i got into, and the fixes i 
found... there are some things which need a bit of time/experience to 
get a camera working here...



THIS mail went to:
Andreas Klemm
Mariusz Woloszyn
Freebsd-gnats-submit
Freebsd Libusb port maintainer
gphoto2 devel maillist


-- 
ANdrei R.
Software Developer
Cronon AG
Regensburg / Germany

------------------------------------[ http://www.goanga.com ]---
      _     _
    o' \.=./ `o   The only people who find what they're looking
       (o o)      for in life are the ones looking for mistakes.
--ooO--(_)--Ooo-------------------------------------------------

Well, I've received only good news related to this patch on the unofficially
rolled libusb-0.1.7_1 port that I distributed (with a call for testers). I
think it's OK to patch this and I will work with the libusb code t get the
patch into the actual package at some point.

Somebody with the commit-bit in Ports, please apply the following patch to
libusb. Note that it will create files/patch-usb.c which needs to be "cvs
add"'ed.

-Jr

diff -urN libusb.orig/Makefile libusb/Makefile
--- libusb.orig/Makefile	Tue Nov 18 22:29:36 2003
+++ libusb/Makefile	Tue Nov 18 22:30:05 2003
@@ -7,6 +7,7 @@
 
 PORTNAME=	libusb
 PORTVERSION=	0.1.7
+PORTREVISION=	1
 CATEGORIES=	devel
 MASTER_SITES=	${MASTER_SITE_SOURCEFORGE}
 MASTER_SITE_SUBDIR=	${PORTNAME}
diff -urN libusb.orig/files/patch-bsd.c libusb/files/patch-bsd.c
--- libusb.orig/files/patch-bsd.c	Wed Dec 31 17:00:00 1969
+++ libusb/files/patch-bsd.c	Tue Nov 18 22:32:25 2003
@@ -0,0 +1,62 @@
+--- bsd.c	Tue Nov  4 12:23:39 2003
++++ bsd-new.c	Tue Nov  4 12:23:21 2003
+@@ -276,7 +276,7 @@
+ int usb_bulk_write(usb_dev_handle *dev, int ep, char *bytes, int size,
+                    int timeout)
+ {
+-  int fd, ret, sent = 0;
++  int fd, ret;
+ 
+   /* Ensure the endpoint address is correct */
+   ep &= ~USB_ENDPOINT_IN;
+@@ -298,8 +298,7 @@
+     USB_ERROR_STR(ret, "error setting timeout: %s",
+                   strerror(errno));
+ 
+-  do {
+-    ret = write(fd, bytes+sent, size-sent);
++    ret = write(fd, bytes, size);
+     if (ret < 0)
+ #if __FreeBSD__
+       USB_ERROR_STR(ret, "error writing to bulk endpoint %s.%d: %s",
+@@ -309,16 +308,13 @@
+                   dev->device->filename, UE_GET_ADDR(ep), strerror(errno));
+ #endif
+ 
+-    sent += ret;
+-  } while(ret > 0 && sent < size);
+-
+-  return sent;
++  return ret;
+ }
+ 
+ int usb_bulk_read(usb_dev_handle *dev, int ep, char *bytes, int size,
+                   int timeout)
+ {
+-  int fd, ret, retrieved = 0, one = 1;
++  int fd, ret, one = 1;
+ 
+   /* Ensure the endpoint address is correct */
+   ep |= USB_ENDPOINT_IN;
+@@ -345,8 +341,7 @@
+     USB_ERROR_STR(ret, "error setting short xfer: %s",
+                   strerror(errno));
+ 
+-  do {
+-    ret = read(fd, bytes+retrieved, size-retrieved);
++    ret = read(fd, bytes, size);
+     if (ret < 0)
+ #if __FreeBSD__
+       USB_ERROR_STR(ret, "error reading from bulk endpoint %s.%d: %s",
+@@ -355,10 +350,8 @@
+       USB_ERROR_STR(ret, "error reading from bulk endpoint %s.%02d: %s",
+                   dev->device->filename, UE_GET_ADDR(ep), strerror(errno));
+ #endif
+-    retrieved += ret;
+-  } while (ret > 0 && retrieved < size);
+ 
+-  return retrieved;
++  return ret;
+ }
+ 
+ int usb_control_msg(usb_dev_handle *dev, int requesttype, int request,



-- 
John & Jennifer Reynolds  johnjen at reynoldsnet.org        www.reynoldsnet.org
Structural / Physical Design - ICG/PNG SCD     jreynold at sedona.ch.intel.com
Running FreeBSD since 2.1.5-RELEASE.               FreeBSD: The Power to Serve!
"Unix is user friendly, it's just particular about the friends it chooses."

Comment 3 andreas 2003-11-25 19:36:09 UTC

State Changed
From-To: open->closed

applied fix from Mariusz Woloszyn <emsi@ipartners.pl> 
has been tested under -current and -stable