enabling a second output makes output unstable

[stefanor]

At DebConf17 we were getting a lot of HDMI resets (or something) on one of our outputs, when both were enabled (+ USB streaming) on v0.0.3-740-g6a8ced6-dirty on opsis.

This lead to 9eb83b4a6645950a13e466893ff20b9683c4ed15 which resolved the problem.

Video was 720p60. Setup was HDMI or VGA -> HDMI into one port of the opsis. With one output to a random NEC projector on the roof above us, over either VGA or HDMI. It seemed to occur pretty reliably no matter what input / outputs we were feeding it.

[mithro]

You were getting unflows when this was happening, right?

Can you get me the debugging output from when it was occurring?

[stefanor]

You were getting underflows when this was happening, right?

Looks like it.

H2U 00:02:08>s
input0:  0x0 (@48.86 MHz) (capturing)
input1:  1280x720 (@74.24 MHz) (capturing)
output0: [email protected] from input1 (underflows: 0)
output1: [email protected] from input1 (underflows: 125)
EDID primary mode:   [email protected]
EDID secondary mode: [email protected]
encoder: 1280x720 @ 24fps from input1 (q: 85)
ddr: 

hardware version info
===============================================
           DNA: 01325c9b471dfc80
           MAC: d8:80:39:ec:15:19

gateware version info
===============================================
      platform: opsis
        target: HDMI2USB
      revision: 4242bb5a89f00437f1580e31e8f762cbcab100d2

firmware version info
===============================================
      platform: opsis
        target: hdmi2usb
    git commit: 4242bb5a89f00437f1580e31e8f762cbcab100d2
    git branch: master
  git describe: v0.0.3-745-g4242bb5-dirty
    git status:
    --
     M ../../../../third_party/litepcie
     M ../../../../third_party/litesata
     M ../../../../third_party/liteusb
    --

         built: Aug 11 2017 04:51:06
        uptime: 00:02:31
-----------------------------------------------