This deadlocking bug is terrible. Replacing the mutex with an error checking mutex using an abort on errors makes the issue simply vanish.
-
-
Yeah, added an abort in both places. Going to figure it out eventually but the long build time makes failed attempts painful.
-
Uhg. Of course Android broke incremental make. Because Android.
-
Incremental builds work but it still takes a long time since it's huge. It's a lot less bad with ninja-backed make, at least.
-
The main issue is that to test production equivalent builds, they need to go through the make dist and release signing process.
-
Could skip that and just have a device using AOSP test keys but it's too painful to keep switching due to having only 1 of each.
-
Have devices set up for running the CTS and it's a PITA to do it over again. Switching keys pretty much requires wiping them.
-
Seems like for (all but final) testing you'd want all the lockdown disabled.
-
Usually develop via userdebug builds signed with the same release keys, to allow switching back and forth with user builds.
- 1 more reply
New conversation -
-
-
And... figured this out. Need to approach signal safety a bit differently. Just going to use a slow, conservative check for now.
-
They have mutexes in signal handlers? Uhg...
-
No but the dynamic object size checks in CopperheadOS need to avoid doing that, and that means picking a hack to handle it.
-
Android sets a sigaltstack by default so it seems to make sense to only do it when the alternate signal stack isn't disabled.
End of conversation
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.