On 15. 08. 19 17:42, Stephen Warren wrote:
On 8/15/19 12:26 AM, Michal Simek wrote:
On 14. 08. 19 17:51, Stephen Warren wrote:
On 8/14/19 12:12 AM, Michal Simek wrote:
On 13. 08. 19 23:21, Stephen Warren wrote:
On 8/1/19 10:48 PM, Michal Simek wrote:
Test is using random command which has own Kconfig symbol CMD_RANDOM which already depends on CMD_MEMORY. That's why replace cmd_memory by cmd_random.
This might not always be true; I think it'd be better to keep the existing dependency list entries and just add CMD_RANDOM:
Can you please tell me when exactly? I didn't see that.
If in the future somebody changes the Kconfig file so that CMD_RANDOM doesn't depend on or select CMD_MEMORY. It may well not be possible right now; I just want to make sure the test doesn't rely on implementation details of CMD_RANDOM/CMD_MEMORY; if it explicitly needs both, it should simply depend on both.
Do you think that we have used this rule in past?
I hope so; I'd consider it a bug in a test if the test used a command but didn't explicitly depend on that command's CONFIG_ flag. We have had such bugs in the past, and fixed one or two of them.
Anyway I understand what you mean but it is questionable if make sense to predict all changes which can happen.
It's more about the test explicitly specifying exactly what it depends on than predicting changes.
I would consider if anybody remove this dependency without fixing test suite as bug which should be fixed. Also tests should detect this and report issues to stop this before patch like this is applied.
True, but people do forget such things.
I have sent a patch. Feel free to fix commit message if needed.
Thanks, Michal