When using the autotest batch files against 2022.021 and 2022.022 Release builds, I'm getting assertions thrown in Stubs.c line 351, GetPtrSize.  Even if I ignore the assertion, the GE exits soon afterward without warning.  Instead of an exitcode of 0 (clean termination of the GE), I see an exitcode of -1073741819.  

I thought this might only be when using the AutoTest batch files where multiple copies of the GE are running simultaneously so I ran just EC2003 MEDC in single-file (one scenario at a time, one copy of GE running at a time) and still see the exitcode of -1073741819 for 50% of the scenarios (8 clean exits, 8 errors) so far.  

Error scenarios:

EC2003 MEDC 3.0 Blue

EC2003 MEDC 5.0 Blue

EC2003 MEDC 7.0 Blue

EC2003 MEDC 8.0 Blue

EC2003 MEDC 9.0 Blue

EC2003 MEDC 12.0 Blue

EC2003 MEDC 2.0 Red

EC2003 MEDC 4.0 Red

EC2003 MEDC 5.0 Red

EC2003 MEDC 6.0 Red

EC2003 MEDC 7.0 Red

EC2003 MEDC 3.0 seems to pretty consistently have a non-zero exitcode.

On 10/19/2022 at 1:58 PM, TonyE said:

EC2003 MEDC 3.0 seems to pretty consistently have a non-zero exitcode.

I removed the XML saving ability in the GE, that didn't help.

 

Then I ran this with the debugger and the VS2022 debugger crashed so didn't get any useful information.  Will try a debug build next.

A debug build of 2022.022 running under the autotest batch files for EC2003 MEDC had zero errors.  Next up is turning off Optimization to see if that sneaks around whatever underlying problem is happening.

I'm guessing us mortals are of no help?

 

On 10/22/2022 at 7:56 PM, donaldseadog said:

I'm guessing us mortals are of no help?

 

At present I don't have a way for you to help.  I logged every memory allocation in EC2003 MEDC 3.0 and didn't have any of my favorite memory bug of allocating the size of a pointer (4 bytes) instead of the size of the structure.  That's rather unfortunate since they are relatively easy to detect when logging all of the allocations.  So on to tougher things to fix.

Note to self: Combat3.c line 605ish around staff_warned with modification dates in 2021 is suspicious, examine it better.

Found and fixed a double-freeing of memory in eff102 (area defense) for a pretty specific situation of aircraft in a non-aircraft formation being shot down by surface to air gunfire.  The test case was indeed EC2003 MEDC 3.0 as Blue where Red ships get within a Blue carrier formation and start shooting down formation air patrols.  Re-running EC2003 autotest batch files now to see how much of an improvement is seen (this is without padding the memory allocations, so the most desirable situation).

And...fixed another memory bug, EC2003 autotest batch was 100% successful after these changes.

2022.023 fix coded