10.3.4 precluding successful use of cameras, me too (fire-i, firewire, 2 cameras, no joy)

[William H. Magill]

On 22 Jul, 2004, at 11:34, Steve Sisak (Support) wrote:
> It acts like a CFRetain()/CFRelease() problem plus a race condition 
> somewhere in QuickTime and happens more often in fast machines -- if 
> one race path wins, no problem, it the other wins theres a reference 
> count underflow and something gets disposed of too soon.

Just as an FYI...

While I was still working at the University of Pennsylvania, (now 
retired), and while Compaq was still an independent company, we 
encountered exactly that kind of problem -- it WAS a race condition, 
and it ONLY showed up on the fast hardware! (We're talking high-end 
Alphas, 64 bit machines, here.)

In that particular case it happened to be a single-threaded process in 
authorization which, on slow machines could "never" have more than one 
caller at at time, but on a fast machine could easily have many.

In those days, we were directly working with their Engineering group, 
so it didn't take too long for us to develop a piece of code which 
showed the failure every time it was run. We sent the code to Compaq, 
and they claimed that it never failed for them!! After a couple of days 
worth of email back and forth, I finally "thought" to ask them -- what 
hardware were they using to test the problem with?  As it turned out, 
Engineering was using a 3 year old box for testing which was a bout 4 
times SLOWER than the hardware we were using. As soon as Engineering 
got the same CPU model that we were using BAM the code failed every 
time!

A bit of further discussion determined that Engineering was in fact 
using ancient desktops for all of their code development and only 
testing things on the departmental sever, which, while newer, was 
itself was 3 years old!

Given Apple's product rollouts -- it would not surprise me to find this 
condition in their Engineering group.

If you have any direct contact with the QuickTime Engineering group, 
suggest to them that they need to test on the latest hardware -- the 
fastest G5.  They may bless you for the suggestion, because Engineering 
always suffers from the "Shoemaker's Children" syndrome. (They're the 
ones who either have holes in their shoes or have no shoes... depending 
on the version of the story.)  It costs money to provide Engineering 
with the latest hardware, and that represents product that doesn't make 
it to the street, so the companies tend to resist upgrading them until 
the last possible excuse has been used up.



T.T.F.N.
William H. Magill
# Beige G3 - Rev A motherboard - 768 Meg
# Flat-panel iMac (2.1) 800MHz - Super Drive - 768 Meg
# PWS433a [Alpha 21164 Rev 7.2 (EV56)- 64 Meg]- Tru64 5.1a
# XP1000  [Alpha EV6]
magill at mcgillsociety.org
magill at acm.org
magill at mac.com