I think it's just zeroing the memory that's going to store a keypress, so it looks fine to me. Also it's been there for 10 years without causing a problem :)

I had some problems with Android 10 concerning storage. I don't know if it's related to this.

There is a long standing bug in the android port that arises once in a while. A monster race pointer becomes null and the game crashes.

"...Google reports a lot of errors in mon_take_hit, get_lore, player_kill_monster. All related in some way to an invalid mon->race pointer..."

As I said, perhaps a rogue memset is wiping a chunk of memory.

Sorry, I didn't know that sizeof could be used without parenthesis.

Maybe this will help...

https://drive.google.com/file/d/1wbl...ew?usp=sharing

That's the full game directory including the save file.

Thank you. I can confirm that this crashes in Win 10, both in the directory you supplied and in a clean copy. Testing continues :mad:

Thank YOU! :)
I loved what I saw of the game so far.

Nick, which compiler do you use for windows ?

If I put a fresh copy of the executable made with mingw (32 bits), the savefile provided by swim doesn't crash.

It's the mingw cross-compiler as picked up from the package, gcc-mingw-w64, in the ubuntu-18.04 runner on Gihub's Actions. configure was run with

Because of Arralen's report of trouble with the Windows executable and Windows 8, have not been using the ubuntu-20.04 runner and its version of mingw.

The latest builds should not have the crash on moving house, thanks to backwardsEric for finding and fixing the bug.

It seems that someone had put some essential code inside an assert(), and then optimised out asserts in release builds...

Excellent! :)
I'm going to give it a try right now!