[U-Boot] mmc tests incorrectly implemented

[Stephen Warren]

On 4/10/19 11:22 AM, Tom Rini wrote:
> On Wed, Apr 10, 2019 at 11:13:32AM -0600, Stephen Warren wrote:
>> On 4/10/19 10:23 AM, Marek Vasut wrote:
>>> On 4/10/19 5:12 PM, Stephen Warren wrote:
>>>
>>> Hi,
>>>
>>> it would be nice if I was CCed on this.
>>
>> Sorry, I didn't drill down in Jenkins/git data to find out where the
>> commits came from; I just had a list of commit descriptions and knew
>> which branch they showed up in,
>>
>>>> I see that some mmc tests have been added to test/py, but I see problems
>>>> with them:
>>>>
>>>> 1) test_mmc_rescan assumes that each entry in env__mmc_rd_configs is a
>>>> separate device that can be rescanned. This isn't actually true; entries
>>>> in that array are intended to drive the mmc read test, and so can point
>>>> at partitions or specific sector numbers.
>>>
>>> Is that documented somewhere ? I assumed they are separate devices and
>>> if you need to test multiple partitions, you will have multiple entries
>>> in this array, one for each device:partition pair.
>>
>> There is an example in a comment at the top of test_mmc_rd.py, since
>> that's what is intended to use this data. Admittedly, it doesn't
>> explicitly spell out that the data array is intended for sole use by
>> that test, but I assumed it was obvious enough since the array was named
>> after the test and only used/mentioned in that one tests.
>>
>>>> Running an mmc rescan test on
>>>> the entire array results in duplicated tests. A new data array should be
>>>> created for different tests.
>>>
>>> I don't have such a usecase, but the fix should be easy to implement.
>>> Can you do that ?
>>
>> Surely you have a use-case for the new test, or you wouldn't have
>> implemented it?
>>
>> I'm afraid I barely have time to keep the test system running. I don't
>> have time to patch up test failures in most cases. The only option I
>> have available is to disable all MMC testing on my boards if that's
>> what's needed to keep the test system going to other tests. Sorry.
> 
> Wait, these are running and failing on your setup?  That wasn't clear,
> sorry.  These showed up as skips on my setup as I haven't tried to
> enable them here yet.

Yes, these new tests run on my system without my doing anything to 
enable them, and the mmc_info test fails. In at least one case, the new 
test runs because it's using the configuration that was intended for the 
mmc_read test.