action #58823

Updated by mkittler almost 2 years ago

In "boot_to_desktop" on ARM the test spams `F2`s on boot to enter some boot menu:

sub tianocore_enter_menu {
# we need to reduce this waiting time as much as possible
while (!check_screen('tianocore-mainmenu', 0, no_wait => 1)) {
send_key 'f2';
sleep 0.1;
(from `` in the opensuse test distribution)

Since the last deployment this is now too slow so the menu isn't entered:

* Last good:
* First bad:

It seems to fail consistently since the first bad run. We've been debugging on openqaworker-arm-2 so some of the subsequent results just failed because of the debugging.

`strace` shows that some IPC writes from isotovideo are quite slow:

[pid 29156] write(4, "{\"json_cmd_token\":\"VEowjWrX\",\"ret\":{\"failed_screens\":[{\"candidates\":[{\"area\":[{\"h\":56,\"result\":\"fail\",\"similarity\":0,\"w\":210,\"x\":804,\"y\":600}],\"error\":1.0,\"needle\":{\"area\":[{\"height\":56,\"margin\":50,\"t"..., 6457775 <unfinished ...>

This slowliness is also visible in the os-autoinst log (compare the last good run the bad run). We also observed other IPC related errors in the strace log:

[pid 30708] write(2, "[2019-10-29T13:07:31.347 UTC] [debug] Backend process died, backend errors are reported below in the following lines:\nsyswrite failed at /usr/lib/os-autoinst/ line 54.\n\tmyjsonrpc::send_js"..., 1875) = 1875

We reverted on openqaworker-arm-2 and the problem was gone:

So something in that change causes a slowdown in the command processing.