[U-Boot] [PATCH] test/py: don't use mmc_rd config for other mmc tests

Stephen Warren swarren at wwwdotorg.org
Wed May 1 22:51:57 UTC 2019


On 4/30/19 10:27 AM, Marek Vasut wrote:
> On 4/30/19 5:29 PM, Stephen Warren wrote:
>> On 4/16/19 4:04 PM, Stephen Warren wrote:
>>> From: Stephen Warren <swarren at nvidia.com>
>>>
>>> Fix test_mmc_dev(), test_mmc_rescan(), test_mmc_info() not to use the
>>> same configuration data that test_mmc_rd() does. Doing so causes the
>>> following issues:
>>>
>>> * The new code uncondtionally expects certain keys to exist in the
>>> configuration data. These keys do not exist in existing configuration
>>> data since they were not previously required, and there was no
>>> notification re: a requirement to add these new keys. This causes test
>>> failures due to thrown exceptions when accessing the non-existent keys.
>>>
>>> * The new tests logically operate on different objects. test_mmc_rd()
>>> operates on ranges of sectors on an MMC device (which may be the entire
>>> set of sectors of a device, or a part of a device), whereas all the new
>>> tests operate solely on entire devices. These are separate things, and
>>> it's entirely likely that the user will wish to runs the two types of
>>> tests on different sets of data; see the example configuration data that
>>> this commit adds. Ideally, the new tests would have been added to a
>>> separate Python file, since they aren' closely related to the existing
>>> tests.
>>>
>>> FIXME: Marek, can you please replace the "???" in this patch with some
>>> reasonable looking data? Thanks.
>>
>> Tom, Marek, any chance of applying this? Right now, every mainline
>> branch is failing 5 tests on every push, which wastes my time having to
>> go through all the logs to manually check that the failures aren't
>> anything new/unknown. Thanks.
> 
> I'm still overloaded, and didn't have time to look into this. But I'm
> really not fond of the duplication here -- now we have two big tables
> describing very much the same thing, which SD/MMC to test .

There's no redundancy; the two different tests are semantically entirely 
different and don't share any configuration. See the patch description 
above for the full details.


More information about the U-Boot mailing list