2 in 3 Crash on startup

[sandinak]

This crash is happening with downloaded versions of 4.12.7, 4.13.0 and 5.0b3. Admittedly I have a HUGE file for this year's show, and while it was mostly stable during runtimes ( we had one crash ), it's not nearly as consistent on startup, and only actually starts about 1 in 3 times. With debugging turned on it's more consistent .. starting 4 in 5 times. The errors look memory related:

Code: Select all

branson@val Showfiles % /Applications/QLC+.app/Contents/MacOS/qlcplus -o ./talesuntold-1.2.qxw

Q Light Controller Plus version 4.13.0
This program is licensed under the terms of the Apache 2.0 license.
Copyright (c) Heikki Junnila (hjunnila@users.sf.net)
Copyright (c) Massimo Callegari (massimocallegari@yahoo.it)

2024-03-28 09:02:09.553 qlcplus[45302:66357815] ApplePersistence=NO
qlcplus(45302,0x700002dd0000) malloc: Heap corruption detected, free list is damaged at 0x6000033e4410
*** Incorrect guard value: 66888611021839
qlcplus(45302,0x700002dd0000) malloc: *** set a breakpoint in malloc_error_break to debug
[1]    45302 abort      /Applications/QLC+.app/Contents/MacOS/qlcplus -o ./talesuntold-1.2.qxw

or

024-03-28 09:03:05.152 qlcplus[45324:66359274] ApplePersistence=NO
qlcplus(45324,0x7000030ef000) malloc: *** error for object 0x60000186cf00: pointer being freed was not allocated
qlcplus(45324,0x7000030ef000) malloc: *** set a breakpoint in malloc_error_break to debug
[1]    45324 abort      /Applications/QLC+.app/Contents/MacOS/qlcplus -o ./talesuntold-1.2.qxw

I haven't been able to identify the issue but have thought it may be related to the ~70 Universes of ArtNET I am running for this show. Looking for any suggestions on debugging this and/or help with solving it. I only see our shows getting bigger

• Showfile can be found here: https://sandsite.org/downloads/talesuntold-1.2.qxw.
• tarball of QLC directory is here: https://sandsite.org/downloads/QLC_.tgz
• this is happening on an iMac 3.8Ghz 8-Core Intel i7 with 128G of memory running Sonoma 14.3.1

PS> With the 4.13.0 release I am now able to build! (THANK YOU SO MUCH) on MacOS 14.3.1 using brew to configure the environment. So if there's debugging I can enable and build here to run a more detailed log than -d .. happy to do so.

PPS> Was playing around in 5.0b3 last night with this file .. taking me a bit but so far really like the improvements. Kudos!

[GGGss]

I want to run your showfile on my Windows 11 NUC PC... but I lack the fixture definitions...

Your project opened without hiccups...
Loading the DMX monitor took quite some time! Hefty

You are using private IP addresses -> [oh darn! QLC+ crashed here also] 172.18.2.** OK
This, I think, is not ok to map the local host to network addresses...

And some use partial transmission, the others full. The reason for that is?

I'm running 4.12.8 GIT. Trying 4.13 now with the DMX monitor open. At Cue 121 it gets heavy
I think I found something. There is a network connection bottleneck. With your settings, I quickly reached 11.7Mb/s. Are your devices capable of these speeds?

[GGGss]

At cue 135, my CPU goes wicked ...
and QLC+ lost all of its controls.

The good news is that I ran your entire show without hiccups...
My setup is above average, though ...

[sandinak]

• the included tarball has the fixture defs, background etc. I use synology sync and then ln -s SynologyDrive/QLC+ ~/libpath/QLC+ across all my machines .. works well ( note do NOT use google drive for this purpose anymore .. it locked us out of QLC at a venue cause it couldn't get network )
• So the local host got default mapped for $reason .. they were mapped to actual IP's .. I'll double check here that is what's in the XML and repost the file.
• Yea .. the host(s) can handle the bandwidth. The stage network is gig router + VLAN'ed and Gig switches. In fact I routinely ran this from an older MBP with 16G memory connected via 802.11a .. no problems. It IS nice that ArtNET is not a very abusive proto. My development desktop is way beefy but still has the crash; so fairly sure it's not memory starvation or similar.
• partial vs full ArtNET depended on the target .. some things handle partial well .. some don't, we've gotten better at identifying them.
• glad you could run it .. was a fun show this year

[mcallegari]

Just for the record: version 4.13.0 has also "standard" transmission mode, which is the Art-Net default method to send data. (basically what MA does)
It is way lighter than continuous transmission (either partial or full)
In this case I would give that a try too.

When I have some time I'll try your project as well.

[sandinak]

Ohh .. I'll test 'standard' when I can get my gear back out .. trailered ATM. I checked and it isn't in the docs yet, and I hadn't noticed when I was testing 4.13, thanks for the heads up! sounds like a good improvement

[GGGss]

Just for the record: version 4.13.0 has also "standard" transmission mode, which is the Art-Net default method to send data. (basically what MA does)
It is way lighter than continuous transmission (either partial or full)
In this case I would give that a try too.

When I have some time I'll try your project as well.

WOW! That's great news—this wasn't mentioned in the release notes. I will have to look into this.

[mcallegari]

yes it was mentioned:

Plugins/ArtNet: add default standard transmission mode as per protocol specifications