#375 DosBox 0.74 crashes when locking Windows workstation

[Mario Rohkrämer]

When I have a game running (not just a DOS prompt), and I lock the Windows workstation (Win+L), DOSBox crashes, and the crash dialog appears on top of the login screen.

Occurs with DOSBox 0.74 on Windows XP Pro SP3, running e.g. GOG "One Unit: Whole Blood", or "Blake Stone" Shareware from the D-Fend Reloaded game packs, but not with "Second Reality" from the D-Fend Reloaded demo packs.

---

DOSBox 0.74, Cpu speed: max 100% cycles, Frameskip 0, Program: BLOOD: dosbox.exe - Fehler in Anwendung

Die Anweisung in "0x00527f32" verweist auf Speicher in "0x0af16010". Der Vorgang
"written" konnte nicht auf dem Speicher durchgeführt werden.

Klicken Sie auf "OK", um das Programm zu beenden.
Klicken Sie auf "Abbrechen", um das Programm zu debuggen.

---

OK Abbrechen

BTW: You lack of current values for the "Group" selection in the Bug Tracker.

[Qbix]

might be the output device that has been selected by gog.com.
I think they select overlay and we have fixed (a few months ago) a bug with that.
So the bug might already have been fixed

[toazz]

Not really, as I said, I also tested the bleeding-edge trunk code.

The case is that, for some reason, my system won't allow a mprotect call to grant execute permissions to malloc'ed memory, as is done at:

src/cpu/core_{dynrec,dyn_x86}/cache.h

even if I completely disable SELinux, the mprotect call will fail (and the code just ignores this, causing dosbox to segfault).

So, I have replaced these malloc/mprotect calls with mmap, and dynamic cpu works now. But I still wonder why the fault only occurs if using opengl or openglnb.

[toazz]

In fact, mprotect succeeds if not using opengl (I just wrote some debugging code). Still wondering why.

[toazz]

oh boy, i should have posted it at 3488013, sorry