Closed jdlcdl closed 1 year ago
This morning, Nick suggested changing "gpu_mem" from 64 to 128 in config.txt on the pi0. This worked for me on pi0, I can no longer reproduce this error.
However, on Raspi-OS I have always used only 16mb for gpu_mem. Changing to 128 and even 256 does NOT resolve the error. I still believe that we should track down exactly where this became a problem. I suspect it's recent, but I'm not sure, It's not a regular habit of mine to load more than one seed at a time... but I've certainly done this in the past and never seen this error before.
I ran into that maybe a month ago but at the time it was because that paste
operation was trying to paste None
instead of image data. It was just a bug in my in-progress code at the time that I resolved. No clue yet what might be causing it this time.
Keith asked me "sticky joystick click?", and this seems to answer everything.
I cannot reproduce the error, on pi0-ss-os nor pi2-raspi-os, without a press-and-hold. If Im careful to quick clickly, I cannot reproduce it at all, while a long-press reproduces it every time.
As of commit beaf4ae which we've been testing for 1 week, I've experienced an exception while one seed was already loaded and trying to create a new seed from image entropy. It's a DireWarning like:
System Error ValueError "Image.py 1505 in paste", "cannot determine region size; use 4 item box"
I've just updated my testing head for Nick's recent merges, to commit
8b5c8af5 9451a9d2244509f and the above error remains.Image.py is part of PIL, so it's a hint, but not so helpful. A backtrace from a Manual Installation pi2 w/ console output looks like:
On the pi0 after regenerating a new pre-release test image in a running pre-built-for-pi0 container, like:
... the same exception is raised in seedsigner-os running on pi0, but the Image.py line number is 1711.
I've tested this a bunch of times now, maybe it has nothing to do with a seed being loaded and more to do with a repetition of create-seed-via-image-entropy. It works the first time, usually not again but sometimes it works without raising this exception.