Fix segfault in model.Tree(procs)

[mikkeloscar]

Fixes crash in model.Tree(procs) which occurs when the list of procs
passed, isn't in the expected order. E.g. if the first element doesn't
have PPID == 0, it would crash writing to the uninitialized
parent.Children slice.

[CLAassistant]

All committers have signed the CLA.

[bradrydzewski]

I think this might need some unit tests, because I'm not sure it will work as desired. A process can have multiple parent nodes where ppid == 0. I don't see anything that clears the list of children. This means if we have 2 parent processes, parent process 2 will have all of parent process 1's children in the list?

I also wonder if this could be solved by sorting? The procs are created with the correct order which has me wondering if the database can return them out of order? I haven't see this issue myself, but I'm mostly tested with sqlite.

[mikkeloscar]

To be completely honest I didn't fully understand the expected output of model.Tree, I just fixed it so it worked the same (as far as I could tell) without crashing.

Your explanation is helpful, is there anywhere else I can look to get a better understanding of how the tree structure should look like?

I'm using postgresql fwiw.

[bradrydzewski]

@mikkeloscar will do. I should have some more time tomorrow. Maybe I can add some initial unit tests and you can build off those when implementing the fix?

[mikkeloscar]

@bradrydzewski sounds good to me! 👍

[bradrydzewski]

can you test if adding an order by proc_pid ASC solves the issue for you?
https://github.com/drone/drone/blob/master/store/datastore/sql/postgres/sql_gen.go#L149:L166

[mikkeloscar]

@bradrydzewski yes, sorting by proc_pid solves the issue for me. I have reverted the other commit and added the changes to the postgres query.

[bradrydzewski]

awesome, thanks! much appreciated