[Bug 37381] client-types.py appears to have a race

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=37381

--- Comment #8 from Simon McVittie <simon.mcvittie at collabora.co.uk> ---
Created attachment 85625
  --> https://bugs.freedesktop.org/attachment.cgi?id=85625&action=edit
Presence: use a sequence number, not a timestamp, to test  "most recent"

This means the client-types test deterministically gets a different
result: it was previously relying on the PC and the phone coming online
near-simultaneously, which resulted in the PC "winning" the comparison
(except when the test ran close to an integer number of
seconds-since-the-epoch, when it failed).

Now, the phone has a more recent "most recently available"
sequence number, so it "wins". This might be considered to make
<https://bugs.freedesktop.org/show_bug.cgi?id=32139> regress,
depending which tie-breaker you consider to be more important.

The change to prefer_higher_priority_resources() illustrates one
effect of this change: the priority never actually matters
unless neither of the resources under consideration has ever
been "available", because the "most recently available"
sequence number is considered more important.

It's not completely clear whether the "most recently available"
sequence number is intended to tie-break between resources that are
away/busy (display the one that went away more recently), or that are
available (display the one that arrived more recently) - at the
moment it does both.

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
You are the assignee for the bug.