Dino explodes memory, CPU usage

Hello,

As of at least commit bb385f247292d3162b27afa7ac48a1bf404c8e37, when attempting to launch Dino, it does not launch and instead consumes all system memory and 100% of a CPU thread's execution time.

Thanks,

Juli

Hi,

"J. Sims" via Bug reports for GNU Guix <bug-guix@gnu.org> writes:

I cannot reproduce on my end with a bare `guix shell dino -- dino`.

What windowing system are you using? Does dino have logs? Have you

tried launching it in a terminal to see if it outputs anything?

Maybe the fact that I've never used it and thus don't have anything

configured might obscure the error.

Best,

--

Josselin Poiret

Hey,

------- Original Message -------

On Thursday, April 27th, 2023 at 16:26, Josselin Poiret <dev@jpoiret.xyz> wrote:

If dino has logs, I do not know where to find them; when launching from the terminal, a newline is emitted and absolutely nothing else before entering the described state. I am using GNOME.

Interestingly, your comment made me curious, so I ran a `guix shell --pure dino -- dino` to see what would happen. It launched fine, so I ran an update of the home profile (where I have it installed) and tried launching dino again - only to reproduce the same problem. I tried your command of `guix shell dino -- dino` - and got the same issue again. However, `guix shell --pure dino -- dino` still works fine. This leads me to believe the issues is another package in my home profile... but which?

I tested all of the packages in my home profile with dino in a pure shell, and it launched fine.

Okay, not packages. Environment variables? I changed some of those recently; let's try reverting them.

Nope, works fine.

The logical next consideration is services, but I'm not sure how to test those without... removing them, reconfiguring home, logging out, logging in... each and every time. It would be nice if you could pass eg `--preserve='^DISPLAY$'` to `guix home container`. Ah well. I'll do the hard thing anyway.

But before I go off to do that, I wanted to send this message with previous debug steps. Especially this interesting tidbit:

Firstly, I misspoke; Dino doesn't consume an entire thread and all the system's memory. It turns out gst-plugin-scanner consumes all the system's memory; the CPU seems to remain unabused.

With this information, I ran `dino --gst-debug-level=6` and found this interesting, repeating pattern:

```

0:00:27.310610920 31838 0x11d8520 LOG GST_PLUGIN_LOADING gstpluginloader.c:1150:exchange_packets: Poll res = 0. 0 bytes pending for write

0:00:27.310655655 31838 0x11d8520 DEBUG GST_POLL gstpoll.c:1241:gst_poll_fd_has_error: 0x13e81e0: fd (fd:25, idx:2) 0

0:00:27.310661435 31838 0x11d8520 DEBUG GST_POLL gstpoll.c:1266:gst_poll_fd_can_read_unlocked: 0x13e81e0: fd (fd:25, idx:2) 0

0:00:27.310667858 31838 0x11d8520 DEBUG GST_POLL gstpoll.c:1195:gst_poll_fd_has_closed: 0x13e81e0: fd (fd:25, idx:2) 0

0:00:27.310675101 31838 0x11d8520 DEBUG GST_POLL gstpoll.c:1414:gst_poll_wait: 0x13e81e0: timeout :0:00:01.000000000

```

This appears to be the action upon which it locks up.

As I said, I'll go test if one of my home services is causing this issue; perhaps this information will be useful in the meantime.

Thanks,

Juli

Problem resolved: apparently the issue was gst-plugins-bad in my system configuration causing problems (which I had hanging around from resolving a website-specific bug in GNOME Web... which I no longer daily-drive anyway).

Thanks for sparking me to actually look into this.

I don't know how to close bugs on debbugs or if this one can be considered resolved, so I'll leave that up to the discretion of more knowledgeable Guix hackers.

-Juli

Hi,

"J. Sims" <jtsims@protonmail.com> writes:

Great to see it's resolved on your end!

You can cc XXXX-done@debbugs.gnu.org instead of XXXX@debbugs.gnu.org

(just like this message), or send a control message to

control@debbugs.gnu.org.

Best,

--

Josselin Poiret