[Intel-gfx] Another behavior of tests from HANG section with enabled command parser

[Chris Wilson]

On Mon, May 18, 2015 at 06:04:36AM +0000, Popov, Pavel E wrote:
> Hi All,
> 
> I observed that some tests from HANG section described in Makefile.in 
> change their behavior when command parser is enabled:
> 
> Tests gem_bad_address starts to fail with error:
>     "CMD: Rejected command 0x10600002 for bitmask 0x00400000..."
> As said Chris this test and gem_bad_blit are out of date.
> 
> Test gem_hang starts to fail with error:
>     "CMD: Rejected master-only command: 0x09000000"
> This is because MI_WAIT_FOR_EVENT marked as master.
> 
> Test gem_non_secure_batch starts to fail with error:
>     "CMD: Rejected register 0x0000203c in command..."
> This looks like an expected behavior for this test which can be handled.
> 
> All "HANG" tests pass on my HSW configuration without command parser.
> But looks like most of them were initially written to fail or hang.
> Am I right that behavior for these tests is unclear at the moment?
> Do you plan to update/fix tests from this section?

They are not really tests, but examples of how you can hang the machine
on particular hardware. There is no expected behaviour for them.
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre