manager crashes on accessing /cover

[jirislaby]

Sometimes, when I go to machine:56741/cover, manager exits. I think, it crashes, as it leaves all qemu instances running.

[dvyukov]

Is there any error message after manager?

[jirislaby]

Not at all, it's like this:
syzkaller> syz-manager -config syzkaller.cfg
2016/02/16 11:21:59 loading corpus...
2016/02/16 11:22:01 loaded 1472 programs
2016/02/16 11:22:01 serving http on http://0.0.0.0:56741
2016/02/16 11:22:01 serving rpc on tcp://127.0.0.1:52034
syzkaller>

I.e. a prompt is shown.

[dvyukov]

Hummm... Can it be an OOM kill? Memory consumption can grow significantly when coverage report is requested.

[jirislaby]

Nope, dmesg is clean. No trace of OOM killer, neither crashes. I ran manager insider gdb, but cannot reproduce yet.

[dvyukov]

Humm... I am out of ideas. Go programs don't exit silently until you say os.Exit (which I don't as far as I see).

If gdb won't show anything useful, let's try to fix profile generation first. Maybe the crash will go away if we fix that...

[jirislaby]

it didn't happen with gdb. Neither it happens w/ debuginfo.

[dvyukov]

OK, I am going to close this. It is not actionable and harmless provided that I add proper diagnosis for missed debug info. Feel free to reopen if you localize the root cause.