New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
RC7 crashes across the board on BPC. #4504
Comments
This is related to the check added in #4500 |
I don't see the |
PR scp-fs2open#4500 added some checks, but accidentally compared `aip->target_objnum` to an object type (although the assertion message correctly retrieved the object type, leading to the bizarre assertion that `OBJ_SHIP`, i.e. 1, was an invalid target object type). This commit makes them all access the `Objects[]` array and retrieve the `.type`, just as the assertion message does (and the pre-scp-fs2open#4500 conditional did). Related to scp-fs2open#4504, in that it's the source of the failed assertions, although I can't be sure it was the cause of the crashes without testing.
That's weird. Delenda Est crashes reliably with that error for me when running either RC7 or the 20-7-2022 nightly on BP 3.0.7 |
I tried both RC7 and the 7-20 nightly on BP 3.0.7, with mission bp2-15, and didn't get the |
I uploaded the current installation. That being said, the RC7 build also crashes in the BP 3.0.3 installation. I should clarify: The crash does not occur on load, but a short time into the mission. |
I still don't see the Do you have any checkpoints saved, and are you starting the mission from a checkpoint? |
PR #4500 added some checks, but accidentally compared `aip->target_objnum` to an object type (although the assertion message correctly retrieved the object type, leading to the bizarre assertion that `OBJ_SHIP`, i.e. 1, was an invalid target object type). This commit makes them all access the `Objects[]` array and retrieve the `.type`, just as the assertion message does (and the pre-#4500 conditional did). Related to #4504, in that it's the source of the failed assertions, although I can't be sure it was the cause of the crashes without testing.
@MoerasGrizzly, is this true for all reported crashes on this issue? |
Yes |
Ok, Please check out RC8 as soon as we get it published. We haven't found the sexp-script-eval error yet, but according to Goober that shouldn't make a crash. We'll demote the priority level to low or close once its confirmed that there are no crashes or freezes related to this issue. |
Got a report on discord that the script crash is still occurring on RC8. |
Script crash is still occuring on RC8 here as well. |
In that case, can you answer my earlier question?
And if you have any checkpoints from that mission, please upload them. |
No. The crash also occurs on Sunglare (BP2-16), which lacks checkpoints alltogether. It consinstently shows up when the Masyaf jumps in. |
This issue has been found with Battle of Neptune using July 25 nightly (scroll up for debug log text; earlier posts may have more info as well). |
I managed to reproduce the issue in the Battle of Neptune (in RC8, instead of the nightly) and I have a coredump that I can send anyone who might want to analyze it. Looking into the debugger, I really can't tell what is happening, but it seems to be related to the Lua stack and memory issues. |
I still have not been able to reproduce this. Tried RC8 and latest master, release and debug, no luck. I hope it isn't caused by #4482. I don't know how it could be, but it fits the time frame. |
@Goober5000 maybe offer @MoerasGrizzly a custom build of master with PR #4482 reverted and see if repro is still possible? @Kiloku maybe try the same sort of build with your Battle of Neptune repro? Were there other recent commits related to the Lua runtime environment? |
copy paste of my discord stream of consiousness rummaging
|
Since the discord conversation mentions errors from within Do we know how many arguments we try to push onto the Lua stack? |
In general, or in this case? Because I think |
My belief now is that the crashes here are a symptom of the problem 4482 was addressing. It added exceptions when out of stack space, lua is running out of stack space, and the game is crashing rather than just continuing into unsafe memory as it appears it would do otherwise. It looks like EvalStringWithReturn leaks stack space, one index per call. Why this hasn't caused bigger problems before I have no idea. I've a PR in the works, hopefully it's a suitable solution. |
Nicely done folks. Looking forward to the next RC. :-) |
Having done quite a bit of playing with the most recent nightlies on BPC , I've detected no more crashes. |
In several War in Heaven missions (that I've found thus far), the RC7 build hard crashes to desktop. These crashes do not occur on the RC6 build.
On Delanda Est (BP2-15) and Sunglare (BP2-16) and the RC7 fast debug build, the following errors are thrown before the hard crash:
In Pawns of a Board of Bone (BP2-14), the following error is thrown:
On "Deals in Shadows" (BP2-13), the following error is thrown:
The text was updated successfully, but these errors were encountered: