Linux Driver Verification

When we improve a model, we sometimes need to check how our new results differs from what we've stored in the KB. For this, not only we'd like to see the categories the new unsafe split to according the KB, but also the unsafes that are missing in this launch, but present in the KB.

Since the KB is somewhat "unsafe-centric", the definition of what should be shown in the "safe zone" is vague. Moreover, there should be a lot of KB unsafes missing in any launch, since the bugs we found get fixed, but KB entries shouldn't go anywhere (a), and all records that relate to the other models, or to the drivers that aren't in the launch at all get shown in the "missing" column (b).

We may approach this problem from two sides. First, we may introduce the concept of "artificial launch": some selected entries from the KB that form a "launch" to compare with (not via the usual comparison interface, though). For that, we may add a Boolean parameter "current" to the KB element, which denotes whether this KB entry is considered for the "Missing" column at all. This helps us with (a), and if the column is editable from the KB admin, we may safely alter it when testing the model on our local copies of KB, thus exploiting the "artificial launch" concept.

Second, we may use the following logic: under the missing column, show only those drivers that match (considering only module_name, rule_model and main) some "current" KB entry that has no other matching drivers in this launch. The developers will be responsible for decreasing the time to update this information by wise selection of "current" KB records.