From: Nathaniel Wesley Filardo <nwfilardo@gmail.com>
Date: Sun, 31 May 2020 23:18:47 +0000 (+0100)
Subject: ws2812 vs gpio: down the AND gate "later"
X-Git-Url: https://hydra-www.ietfng.org/gitweb/?a=commitdiff_plain;h=b3c545be719311cf8a1d5ce035efa8c13cf3c4b5;p=acmetensortoys-esp-lua_lamp

ws2812 vs gpio: down the AND gate "later"

Right now, the ws2812 driver doesn't wait for the bit stream to return
control to Lua, so rather than immediately kick the AND gate off, set up
a timer to do it "soon".  This is not really the right fix, since it
leaves the AND gate high for a long while, but it works well enough I
suppose.
---

diff --git a/init2.lua b/init2.lua
index df28251..44f7527 100644
--- a/init2.lua
+++ b/init2.lua
@@ -35,6 +35,8 @@ end)
 ws2812.init(ws2812.MODE_SINGLE)     -- uses GPIO2
 i2c.setup(0,2,1,i2c.SLOW)           -- init i2c as per silk screen (GPIO4, GPIO5)
 
+local inhtmr = tmr.create()
+
 -- and now we get to the lamp stuff
 remotefb = ws2812.newBuffer(32,3)
 ledfb = remotefb -- points at whichever buffer is appropriate to draw
@@ -60,7 +62,11 @@ function dodraw()
     else
       gpio.write(3,gpio.HIGH) ws2812.write(ledfb)
     end
-    gpio.write(3,gpio.LOW)
+    -- Kick the inhibit timer to turn off the AND gate in 2 msec.  That should
+    -- be ample time to drain the serial UART.  If we get here again, we'll end
+    -- up resetting the timer's firing time, making this something like a
+    -- watchdog.  Hopefully we won't crash in the interim.
+    inhtmr:alarm(2,tmr.ALARM_SINGLE,function() gpio.write(3,gpio.LOW) end)
   end
 end