Comment 159 for bug 226279

Again, my current machine is an iMac Core2Duo circa October 2007.

I have resolved my suspend/resume issue thanks to the Hardy proposed -19 kernel, the EnvNG utility, and a web page (URL below).
For me, all 3 were necessary. Now, I can report that:

   -12 worked in the past (we all know that)
   -17 failed as reported by lots of folks
   -18 ditto
   -19 (proposed) fixes my iMac suspend/resume issues in -17 and -18

The kernel needed a cooperative video driver. Given that Intel-ish iMacs require the proprietary ATI driver (fglxr), there is no automatic way to know that your current driver will not work with newer kernels 100% (E.g. resume after suspend fails while everything else works). There are similar issues with nVidia.

Lesson learned:
1 - Run EnvyNG after every kernel upgrade *until* (hopefully) fglxr (or an equivalent) becomes part of the ubuntu family.
2 - Check ACPI configuration to see if a driver is stubborn and not letting go and force it to let go.

Packages needed envng-core, envng-gtk [or envng-qt]
Quote: "install the ATI or the NVIDIA driver EnvyNG is an application written in Python which will download the latest ATI or NVIDIA driver or the Legacy driver (for older cards) (according to the model of your card) from ATI or Nvidia's website and set it up for you handling dependencies (compilers, OpenGL,
etc.) which are required in order to build and use the driver."

ACPI configuration: I also needed the procedure at

   http://ubuntulinuxtipstricks.blogspot.com/2008/03/ati-nvidia-resume-good-news-bad-news.html

which, in a nutshell, contains a procedure to have ACPI to unload the ATI (fglxr) driver on suspend and reload on resume. Intel iMac users, do not set SAVE_VIDEO_PCI_STATE=true in /etc/default/acpi-support or you will probably be sorry [E.g. lock up].

One minor annoyance remains in ACPI. LOCK_SCREEN=false in /etc/default/acpi-support (or just commenting out the original LOCK_SCREEN=true) doesn't seem to work. It still locks the screen no matter what the LOCK_SCREEN condition is.

Sorry if I unnecessarily bothered any of the kernel folks. Our time is precious.

-Richard,
Dating myself: BSD 4.1 driver development on Vaxen 30 years ago just because it was fun