We had reports [on github](https://github.com/mozilla-mobile/fenix/issues/16469.) about display corruption on a Mali-G77 device. This was filed on bugzilla as bug 1676474. I didn't have a device which could reproduce, but with a lot of help from the reporter we managed to establish that it was due to partial present, and therefore disabled partial present on Mali-Gxx devices. (As far as we're aware it only affects G77, but we disabled on all Gxx just to be safe.) This bug tracks figuring out the problem and fixing/working around it, then re-enabling the feature. The user reported the bug appeared, then went away, then re-appeared. Using mozregression I have found the following ranges: [Bug first appeared](https://hg.mozilla.org/integration/autoland/pushloghtml?fromchange=455ff5647429040ea927ab6f50855afb93a156e2&tochange=f0df0bb8123db84dc77fb7d32fdf9e5045e4f0e7) - Bug 1675159. We suspected this at the time based on the dates. It is a refactoring of related code. It shouldn't *in theory* have caused this, but here we are. As we said at the time, it's likely a subtle driver bug which this change uncovered by affecting the orderings of GL calls. [Bug went away](https://hg.mozilla.org/integration/autoland/pushloghtml?fromchange=0035088d5fa64a0d1ead4bc1925e31b8719ffdee&tochange=5f5c269d15d6e521b15243b5014a65965df0cc08) - Likely bug 1661528. Again, this doesn't really make much sense, although it is consistent with... [Bug reappeared](https://hg.mozilla.org/integration/autoland/pushloghtml?fromchange=d22dc65be538930b205704bb8026dd7dfd51a833&tochange=98c926d9289f5c45565a3a157971b46f98210133) - Bug 1676390. The bug went away when I landed a change which caused us to use persistently mapped buffers, and I think most importantly, call eglFenceSync midway through the frame (after texture uploads). The bug then reappeared when I moved the glFenceSync call to the end of the frame.
Bug 1685276 Comment 0 Edit History
Note: The actual edited comment in the bug view page will always show the original commenter’s name and original timestamp.
We had reports [on github](https://github.com/mozilla-mobile/fenix/issues/16469.) about display corruption on a Mali-G77 device. This was filed on bugzilla as bug 1676474. I didn't have a device which could reproduce, but with a lot of help from the reporter we managed to establish that it was due to partial present, and therefore disabled partial present on Mali-Gxx devices. (As far as we're aware it only affects G77, but we disabled on all Gxx just to be safe.) This bug tracks figuring out the problem and fixing/working around it, then re-enabling the feature. The user reported the bug appeared, then went away, then re-appeared. Using mozregression I have found the following ranges: [Bug first appeared](https://hg.mozilla.org/integration/autoland/pushloghtml?fromchange=455ff5647429040ea927ab6f50855afb93a156e2&tochange=f0df0bb8123db84dc77fb7d32fdf9e5045e4f0e7) - Bug 1675159. We suspected this at the time based on the dates. It is a refactoring of related code. It shouldn't *in theory* have caused this, but here we are. As we said at the time, it's likely a subtle driver bug which this change uncovered by affecting the orderings of GL calls. [Bug went away](https://hg.mozilla.org/integration/autoland/pushloghtml?fromchange=0035088d5fa64a0d1ead4bc1925e31b8719ffdee&tochange=5f5c269d15d6e521b15243b5014a65965df0cc08) - Likely bug 1661528. Again, this doesn't really make much sense, although it is consistent with... [Bug reappeared](https://hg.mozilla.org/integration/autoland/pushloghtml?fromchange=d22dc65be538930b205704bb8026dd7dfd51a833&tochange=98c926d9289f5c45565a3a157971b46f98210133) - Bug 1676390. The bug went away when I landed a change which caused us to use persistently mapped buffers, and I think most importantly, call glFenceSync midway through the frame (after texture uploads). The bug then reappeared when I moved the glFenceSync call to the end of the frame.
We had reports [on github](https://github.com/mozilla-mobile/fenix/issues/16469) about display corruption on a Mali-G77 device. This was filed on bugzilla as bug 1676474. I didn't have a device which could reproduce, but with a lot of help from the reporter we managed to establish that it was due to partial present, and therefore disabled partial present on Mali-Gxx devices. (As far as we're aware it only affects G77, but we disabled on all Gxx just to be safe.) This bug tracks figuring out the problem and fixing/working around it, then re-enabling the feature. The user reported the bug appeared, then went away, then re-appeared. Using mozregression I have found the following ranges: [Bug first appeared](https://hg.mozilla.org/integration/autoland/pushloghtml?fromchange=455ff5647429040ea927ab6f50855afb93a156e2&tochange=f0df0bb8123db84dc77fb7d32fdf9e5045e4f0e7) - Bug 1675159. We suspected this at the time based on the dates. It is a refactoring of related code. It shouldn't *in theory* have caused this, but here we are. As we said at the time, it's likely a subtle driver bug which this change uncovered by affecting the orderings of GL calls. [Bug went away](https://hg.mozilla.org/integration/autoland/pushloghtml?fromchange=0035088d5fa64a0d1ead4bc1925e31b8719ffdee&tochange=5f5c269d15d6e521b15243b5014a65965df0cc08) - Likely bug 1661528. Again, this doesn't really make much sense, although it is consistent with... [Bug reappeared](https://hg.mozilla.org/integration/autoland/pushloghtml?fromchange=d22dc65be538930b205704bb8026dd7dfd51a833&tochange=98c926d9289f5c45565a3a157971b46f98210133) - Bug 1676390. The bug went away when I landed a change which caused us to use persistently mapped buffers, and I think most importantly, call glFenceSync midway through the frame (after texture uploads). The bug then reappeared when I moved the glFenceSync call to the end of the frame.