From: Michael Ginis (mginis++at++xig.com)
Date: 03/20/2002 11:28:23
Dear Performer users/devlopers,
There have been reports that the Linux version of perfly distributed
with Performer 2.5 segfaults almost immediately after startup. The
problem is manifested as a core dump from one of the perfly threads,
leaving the other thread (which appears to do the rendering) running
in the background but inoperative. Sometimes the the perfly GUI is
left visible on the screen, in other cases it never gets that far.
After invesitagting, QA found that this happens only on certain Linux
OS distributions with our Accelerated-X servers, which has our own
accelerated OpenGL implementation and different extensions than
XFree86/NVIDIA OpenGL implementations. By distribution I mean
different versions of Slackware, Mandrake, etc...
On most distributions perfly works with Accelerated-X, however on
those distributions that it does not work we have not been able to
isolate anythong wrong with Accelerated-X that would cause the crash,
or anything special about those distributions which would explain why
it crashes. We managed to get a stack trace out of perfly, which
showed the last call on the stack to be 'pfChannel::computeStress'
initiated by a 'pfFrame' call.
We tried to reproduce the problem with XFree86/NVIDIA servers, but
stock servers did not show this problem on the same distributions
where perfly was crashing with Accelerated-X. However, one thing
different between XFree86/NVIDIA and Accelerated-X is that Performer
tries to query the refresh rate using the XFree86 VidMode extension,
and perfly warns that this extension is missing with Accelerated-X.
(Accelerated-X does not support this extension, because we have our
own extension and library to query/set resolutions and rates, XiGMisc
and XME).
We removed the XFree86 VidMode extension from the XFree86/NVIDIA
server and then perfly started to crash for XFree86/NVIDIA as well.
So we think that there may be a problem in 'libpf', possibly related
to an uninitialized variable or reading the refresh rate. We haven't
been able to find sources for 'libpf', so we haven't been able to
investigate further. We'd like find the problem, since a number of
our customers use Performer. Thanks for any ideas or help.
Sincerely,
-- Mike Ginis mginis++at++xig.com http://www.xig.com ------------------------------------------------------------------------
This archive was generated by hypermail 2b29 : Wed Mar 20 2002 - 11:29:05 PST