[Stefan Westerfeld <email@example.com>] Re: Not entirely fixed. - big-endian arts issues
--- Begin Message ---
On Wed, Oct 31, 2001 at 08:50:48PM -0700, Ivan E. Moore II wrote:
> Response on the arts patch you supplied.
> On Wed, Oct 31, 2001 at 09:04:51AM -0500, John Goerzen wrote:
> > There seems to have been progress, but it's not entirely there yet.
> > The arts daemon no longer complains about endianness, but still
> > doesn't work. It sometimes complains of a "sound server fatal error:
> > CPU overload" when trying to play a sound or when starting up the
> > server. It still doesn't actually play anything. I did make sure to
> > try forcing it to 44100 Hz and 16-bit (formats supported by the
> > hardware), but this did not help at all. I'm thinking there is still
> > an upstream bug.
Ok - thanks for the feedback, I am closing the bug now (and putting the fix
in the KDE_2_2_BRANCH). To the other issue, CPU overload:
This is probably not an aRts bug. Basically, the kernel driver can cause artsd
to fail in two ways:
a) select for writing on audio FD wakes up too early (nothing to do)
b) GETOSPACE ioctl implemented wrong (lies about the actual free space)
If you use strace, you will see that artsd is woken up again and again by
select() far too often. I know that the problem exists for a few kernel drivers
(like i810), but aRts isn't the place to fix it. I also know that other
programs that use the kernel drivers differently don't experience trouble
(i.e. xmms), but still its a kernel bug. If you need more information, let me
know. Please also consider reporting it to the author of the kernel driver.
-* Stefan Westerfeld, firstname.lastname@example.org (PGP!), Hamburg/Germany
KDE Developer, project infos at http://space.twc.de/~stefan/kde *-
--- End Message ---
John Goerzen <email@example.com> GPG: 0x8A1D9A1F www.complete.org