Merge lp:~vanvugt/mir/fix-1665802 into lp:mir
Proposed by
Daniel van Vugt
Status: | Rejected |
---|---|
Rejected by: | Daniel van Vugt |
Proposed branch: | lp:~vanvugt/mir/fix-1665802 |
Merge into: | lp:mir |
Diff against target: |
163 lines (+105/-5) 3 files modified
src/client/frame_clock.cpp (+10/-1) src/client/frame_clock.h (+1/-0) tests/unit-tests/client/test_frame_clock.cpp (+94/-4) |
To merge this branch: | bzr merge lp:~vanvugt/mir/fix-1665802 |
Related bugs: |
Reviewer | Review Type | Date Requested | Status |
---|---|---|---|
Daniel van Vugt | Disapprove | ||
Review via email: mp+318565@code.launchpad.net |
Commit message
Proof of concept fix. This might be enough, or maybe not.
To post a comment you must log in.
Unmerged revisions
- 4068. By Daniel van Vugt
-
Tidy up the test
- 4067. By Daniel van Vugt
-
Correct and clean up the fix. Tests all pass
- 4066. By Daniel van Vugt
-
Expand the test more to fail due to the latching
- 4065. By Daniel van Vugt
-
Enable the fix (and debug output). Tests pass but the flag is latching.
Needs a better test - 4064. By Daniel van Vugt
-
Disable the fix and rewrite the regression test
- 4063. By Daniel van Vugt
-
Merge latest trunk
- 4062. By Daniel van Vugt
-
Expand the test and leave in debug output
- 4061. By Daniel van Vugt
-
Fix the fix. Now all tests pass.
- 4060. By Daniel van Vugt
-
Try to fix the bug, and bad expectations in older tests
- 4059. By Daniel van Vugt
-
Write a detailed regression test (presently failing)
I'm hesitant to endorse this still, because it reduces our ability to catch up in almost- full-framerate scenarios across the board (instead opting to sleep and hope that's enough to give the server more GPU time). I feel this is something we shouldn't be "fixing" because it's not a Mir bug and more important devices (like everything other than Freedreno) could suffer slightly if we land this.
I'll look at other workarounds still that don't make the same sacrifice (although may be much less clean in code than this).