Hi !
Julien BLACHE a écrit :
Frederic Peters <[EMAIL PROTECTED]> wrote:
Hi Frederic,
The problem lies in libusb, it increased its block size from 4k to 16k
libusb strikes again \o/
Actually I am still convinced the problem is not in libusb. This size is
only the block size used to dialog with the kernel. It has nothing to do
with the USB device itself.
It means either the problem is in the kernel (hello Greg), or in
libsane/libgphoto2 itself.
to improve performance of USB 2 in 0.1.10 and this broke libgphoto2[1],
so I got to learn about that stuff and tilted when Jerome told me
about his problem.
Also, I talked with Aurelien Jarno before I had a patch for libgphoto2
and he agreed to upload libusb 0.1.10 with 4k blocks to t-p-u, he did
it but I told release managers it was no longer necessary.
Nice. So now, Aurélien is away for a week, and I have the choicehe
between asking for libusb to be pushed in Sarge (provided the version
in Sid is fixed wrt block size) or rush a new upload of SANE into
Sarge.
Actually, I found a place when I can reach the GSM network, and so read
I can read my mails, even if I am not really reactive.
I would prefer I change in SANE instead of libusb, so that other
applications using libusb won't suffer to performance losses.
Well, please find the patch attached, it is based on the one used in
libgphoto2.
Looks like it is more a workaround than a fix, but it's already better
than nothing. Thanks for it.
Oh wait, I am thinking of something about the timeout value. As the
usb_bulk_read function is now trying to read more data, the time need to
do that is a bit more longer. Maybe the timeout value has to be
incremented. That would explain why your patch is working, but it
wouldn't explain why reducing the blocksize to 4 kB is working.
Bye,
Aurelien
--
.''`. Aurelien Jarno | GPG: 1024D/F1BCDB73
: :' : Debian GNU/Linux developer | Electrical Engineer
`. `' [EMAIL PROTECTED] | [EMAIL PROTECTED]
`- people.debian.org/~aurel32 | www.aurel32.net
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
