Kernel prepatch 3.11-rc4

We don't often use a compare as a function parameter in the kernel so it's not a problem in practice. We don't like it when people use assignments as a parameter. If I spot an assignment like that in a patch, I will tell the submitter to fix it in a follow on patch. But now I will write a Smatch check to spot it automatically so I can complain every time.

I have reviewed the kernel patches so far this year. Out of 39,745 total patches, there are 21 patches where only a '=' character changed. There is only one '=' vs '==' typo and this ARM bug will be the second bug like that this year. The remaining 20 bugs from the list are almost all off-by-one bugs.

The '=' vs '==' patch was: 9e48854c58 "drm/tilcdc: Fix an incorrect condition". It was found with by checkpatch.pl. Checkpatch is a mandatory step for patches and it reflects a problem in the process that it was skipped for this patch.

It should also have been caught by GCC except the authors of that driver put extra parenthesis around everything as if they are programming in LISP. In the kernel, LISP-isms are considered bad style. We don't have an automated process to complain about this, but I do try to educate submitters to drivers/staging/ about this.

The tilcdc bug would have been caught by Smatch as well but it is an ARM thing. The new CONFIG_COMPILE_TEST might help with this in the future.

Apparently the driver was tested, but I have to think that it wasn't tested very well. The maintainer, Rob Clark, did not have a board to test it on at the time.