Deep Sleep: Fix suspend script crash #595
Conversation
|
We also have the retry mechanism on the caller side: https://github.com/LoveRetro/NextUI/blob/main/workspace/all/common/api.c#L3799 |
|
I mulled this one over, if we do that we should fix the fallback implementation PLAT_deepSleep so it doesn't cause the system to fully power off on a single failure. |
Yeah, sounds good - lets do it that way. PLAT_deepSleep already checks return value, just needs a loop. |
|
Looks clean! |
|
The fix as it had been initially designed had some unacceptable cases of non-responsiveness to user input. 4117654 introduces a new helper, waitforwake.elf, that works like Also fixed: a really frustrating, hard-to-debug edge case where the system goes to sleep normally but returns non-zero on wake. When the user presses the power button to wake, the script thinks that this was a failed sleep attempt and puts the system back to sleep. In my testing I've had cases where I've tried 4 times to wake the console and it failed 4 times in a row. I've wasted enough of my life trying to pin this one down so I'm just going to check how long the console was asleep for and assume it was successful if longer than 5 seconds. This is a number of big changes so open to feedback as always. |
|
For something that ideally should just be |
|
Sure, the question is what to leave in... I think some amount of power button unresponsiveness is ok but the bug where the system falls back asleep immediately after waking is unacceptable. |
Fix compile issues for x86-64 devices
… from other devices (LoveRetro#580) * State_read: Dedupe unnecessary branch rzfile already handles uncompressed files for us * do not display stale errno here * Save States: improve compatibility with save states imported from other devices
fix: nextui hanging when launching games from folder auto-launch
…Retro#615) * Fix: improve PLAT_bluetooth* to cope with API changes between bluez versions * Fix: Use MAC address as map key to prevent collisions with duplicate/empty device names
LoveRetro#616) Fix: wifi fails to come up after reboot, if bluetooth is also enabled and inits at the same time
* Fix: Clock pak deployment * Fix: clock rendering on wrong surface format
Deep sleep: revert excess changes, remove waitforwake, etc.
DrFlarp
force-pushed
the
recover-from-bad-suspend
branch
from
6f138ba to
cfcb84f
Compare
4 participants
If kernel returns non-zero on suspend, handle the error properly and retry without exiting the script.
This avoids having to constantly restart network services. The user will not be able to wake the device in the time that it is retrying the deep sleep process (around 6 seconds on a Brick with Wifi turned on)
This addresses #594