For those that carve their own kernels, I think I just got a clue about an NVIDIA driver install failure that doesn't leave a damned clue behind when it fails to install. A failure that has caused me to tear out a considerable amount of hair I can't afford to lose at my age (76).
I finally, by unpacking the -custom.run you get from running it with the --add-this-kernel option, then stepping through the install failure log and comparing it, I found that when it fails, it actually fails at the "insmod ./kernel/nvidia.ko" function with "NVIDIA: no symbol version for module_layout" in the dmesg output. nvidia of course, in their infinite wisdom, simply says it died, with no clue why actually output to the log.
So I went back to a "make xconfig" on my shiny new 2.6.37 and re-read the help for module versioning about 3 times before I grokked that what it said was effectively the opposite of what I was reading, so I turned it off & reran my kernel build/install script.
Then rebooted to level 3, went through the --add-this-kernel thing again, and the install then worked like a champ, so I'm on kernel 2.6.37, using the nvidia latest .29 driver.
Moral, turn off the module versioning, it doesn't mean what it says.
"There are four boxes to be used in defense of liberty:
soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
What you call data processing when people are so disgusted with
it they won't let it be discussed in their presence.