Inner Space is Crashing on exiting to character select

Isboxer.txt

(823.95 KiB) Downloaded 16 times

I broadcast a command to camp my characters. 18 go to the character select screen. I get the following message for the remaining characters. Those seem to remain in game. The error message is as follows:


Inner Space has experienced an unrecoverable problem

C:\Program Files (x86)\Microsoft\Edge\Application\msedge.exe unable
to link imported msedge_elf.dll (-6)

After exiting the error dialog box for any character instance above 18, the remaining characters go to the character select screen. This seems to happen for anything more than 18 characters.


I've attached the export of my isboxer file without the station names. It's possible something the problem is there. I have a single 49 inch screen divided in isboxer as two displays. The right display has the select for an instance and the left display is the currently active instance. I've use this configuration for 18, 24, and 32 characters.

There is also a loggin problem. Random characters get stuck in the server select screen. They stay there for a random amount of time and mostly do finally get selected. But sometimes I need to select the character set a second time.

On an occasional basis, I get a crash report. Most of the time it disappears before I have a chance to fill it out. Below is a dropbox link to a picture of the crash report. I couldn't finish the report before the cursor could not enter the form to press the send button. Below is a link to a picture of the report:


https://www.dropbox.com/scl/fo/kd5kknl6 ... fddgu&dl=0

[bob]

This might not resolve anything at all, or it might help a little. Looking through the dump there is possibly something that needs an IS update. That said.

It would be interesting to know if some of your issues are related to the CPU core limitations.

• you have a 64 thread CPU.
• you have 1 thread allocated to each game
• with the way core allocation works, it is probable that some of your games are on the primary thread and some are on the secondary thread (HT in Intel CPUs).
• 1 thread is going to be very restrictive for EQ these days considering they have just updated to a more heavy required DXAPI and it is still a bit in flux.
• The toolkit is limited to assigning 32 threads currently.

Personally, if I was not going to use a CPU strategy to just let Windows figure out who needed what (there is an option in the CPU strategy Wizard), I would at least assign 4 threads to each game. These can overlap, so for example Slots 1 - 4 all get assigned threads 1 - 4. Just so there is some flexibility.

However, I would recommend trying this first and see if it improves anything (as essentially it should let Windows use all 64 threads if it wants to).

I have more information on the dumps. Innerspace must be crashing. When I click on the ICON in the hidden icon selection, it disappears The task manager cannot find a Inner Space Uplink in the background processes. When I click on the right screen for the instance selection, it disappears. There is one instance in the instance selection which is frozen. I know the character is not online. The everquest icon in the task tray showing all the instances including the instance shown as frozen. When I select from the list provided by the everquest Icon on the task bar nothing is selected. The other characters show on the screen when selected by this method.

I then rerun innerspace. No icon is shown in the task bar or hidden icons. Inner Space Uplink is now found in the task manger as a background process. I do a dump on this instance of Inner Space Uplink. The dropbox link will show this second dump. I'm not sure of the utility of a dump on this second instance of innerspace. The icon doesn't show so something funny is going on.

It required over 12 hours of time for the problem to show up this time. I'm not sure if it is related to how actively I'm using the difference instances. The less activity the more likely the problem shows up.

Let me know what I can do to dump when innerspace crashes.

I'll try the CPU strategy change and let you know the results. The definition of a core is a little slippery. I have 32 cores or 64 threads. Isboxer is assigning cores. AMD EPYC 7513.

[bob]

The definition of a core is a little slippery. I have 32 cores or 64 threads. ISboxer is assigning cores. AMD EPYC 7513.

It is indeed slippery (fixed my earlier post). The affinity setup assigns threads because the OS sees threads (calls them "logical processors") and assigns them a number (1 to 64), which is then used to set affinity. ISBoxer is currently limited to assigning 32 threads.
To be fair, I don't have an AMD setup, so it is possible that the API calls are somehow only returning the cores and setting affinity then sets per core, but I doubt it. The API's generally use consistent behaviour between AMD/Intel CPUs.



RE: Crashes.
Yes, there did look to be something to do with InnerSpace in the first crash dump. That has been logged.

I'm confused.

I have a laptop with an Intel 8 core, i9-9880H, processor. The Isboxer 42 CPU Stategy module says I have 16 cores and assigns 2 cores (threads) to each character slot. Which is correct. The 32 core AMD processor shows as 32 cores in the CPU Strategy module of Isboxer 42. It assigns 1 core to each character slot which is correct. It seems AMD and Intel processors are treated differently. What the API is doing with this information, I don't have a clue.

[bob]

***I might have to walk back my edits.

Anyway, it is recommended that for a high core/thread count CPU that you run letting Windows figure out what core each game can run on. That way you are not limiting the game, and it can be scheduled wherever is needed if some particular core/thread happens to be busy doing other things, like dealing input devices or network traffic etc.



** if you are really interested and want to satisfy my curiosity, you can run the attached vbscript (you'll need to rename so the extension is .vbs). Post the output, and then I'll re-evaluate what I wrote based upon that (I don't have an AMD CPU, let alone an EPYC, to test with).