It's possible to get the reftest harness into an infinite drawing loop if we ever return DrawResult::BAD_IMAGE, because DrawResult::BAD_IMAGE is unrecoverable, but we'll keep trying to draw because ShouldInvalidateToSyncDecodeImages() keeps returning true. Let's make it return false in this scenario. This is currently causing the 944353.jpg crashtest to fail intermittently. Once downscale-during-decode is enabled, it becomes permafail, so we need to land this first.
Created attachment 8577792 [details] [diff] [review] Give up on sync decoding if we get DrawResult::BAD_IMAGE Here's the patch. It's a one-liner.
Attachment #8577792 - Flags: review?(tnikkel)
Attachment #8577792 - Flags: review?(tnikkel) → review+
Thanks for the review! Pushed: https://hg.mozilla.org/integration/mozilla-inbound/rev/c82354a9cf84
Status: NEW → RESOLVED
Last Resolved: 3 years ago
status-firefox39: --- → fixed
Resolution: --- → FIXED
Target Milestone: --- → mozilla39
You need to log in before you can comment on or make changes to this bug.