Linux Driver Verification

I'd like to see on the index page only the summaries. I.e. the columns like "tagged unsafes", "untagged unsafes", and "tagged non-unsafes", each linking to the proper auxillary page. I don't think that we really need tags on the front page.

Eugene Novikov wrote:

I just want to imagine what will happen if we will place KB tags in the way like problems at the beginning of the main statistics page near or under 'Unsafe' column...

Under UNSAFE column I expected to see KB verdicts only. KB tags should be in separate columns.

May be they should be hidden by default?

No, they should be available in our "default" profile, but they may be hidden in some of other predefined profiles.

Moreover after all this statistics (both for verdict and for tags) should become links on corresponding pages with detailed information. For instance, show all unsafes from a given task with verdict 'True positive' and so on.

That is definitely true.

Okay, so at the meeting we had today, we agreed on the following initial design. After the "safe-unsafe-unknown" columns, there is a "cluster" of columns, just like for the other instruments, with such columns as "untagged unsafes", "false positives", "true positives", etc, and with "tags" laid out just like problems in the other instruments.

A potentially useful column "Missing unsafes" is discussed in feature #1708

Accepted on demo 18. Will be closed as soon as KB is merged to master.

BTW, I just have noticed, that the numbers of unknowns for Index and for Unsafes pages don't coincide. For Unsafes this number is the number of unknowns + the number of non analyzed unsafes, for Index it's just the number of unknowns.

Sorry, I have read incorrectly (thought that explicit unknowns aren't counted).

I understand that it's too easy to implement: the number of implicit Unknowns among Unsafes can be calculated as the total number of Unsafes minus the number of analyzed Unsafes (i.e. either True Positives, False Positives or Unknowns). We can do this directly in statistics visualizer, we shouldn't extend KB, KB cache, etc.

Evgeny Novikov wrote:

I understand that it's too easy to implement: the number of implicit Unknowns among Unsafes can be calculated as the total number of Unsafes minus the number of analyzed Unsafes (i.e. either True Positives, False Positives or Unknowns). We can do this directly in statistics visualizer, we shouldn't extend KB, KB cache, etc.

Alexey suggested that we also shouldn't forget about a detailed list of Unknowns among Unsafes (it should include both explicit and implicit Unknowns).

Evgeny Novikov wrote:

BTW, may be rename statistics server with LDV analytics server at the public level (i.e. change the title "Linux driver verification statistics" with "Linux driver verification analytics server"). After coming Knowledge Base, our server provides users not only with a statistics representation but also with some kind of analytics (I guess that analytics includes statistics).

I like 'LDV Analytics Server'/'Linux Driver Verification Analytics Server' name.

Works on QA now, so the issue could be closed.

Alexey, please verify and close the issue if all is okay.