PLH::BreakPointHook crashes when exception handled about polyhook_2_0 HOT 3 CLOSED

Resolved :)

ee07729

You will need to wait for microsoft to merge this PR before you can use vcpkg to install polyhook. Though you can mirror these changes locally in your own portfile and uninstall then reinstall polyhook via vcpkg to immediately get the update.

microsoft/vcpkg#15537

from polyhook_2_0.

Hi, thanks for reporting this and providing such a clean reproduction case! I've identified the bug, in a recent commit I introduced the ability to provide a custom filter callback so users gain more control over how the root exception handler works. This handler is invoked here:

Unfortunately I somehow forgot the case of when users do not use this. What occurs is that an empty std::function eventually gets invoked which is a big no-no and that then throws a C++ exception. But this exception isn't obvious because we threw it from an exception handler, which tries to handle it by calling into the exception handler again, which throws another one! This cycle of throwing and catching occurs forever until the stack overflows and eventually is caught by _chkstack which crashes the program with an invalid read. Fun!

I will fix this soon. In the meantime as a workaround, please try to define an empty handler with this logic:

PLH::AVehHook::EventException() += [](EXCEPTION_POINTERS* exInfo, DWORD* returnCode) noexcept {
        DWORD exCode = exInfo->ExceptionRecord->ExceptionCode;

        if (exCode == EXCEPTION_BREAKPOINT) {
            printf("PolyHook2 about to handle breakpoint exception at %I64X\n", exInfo->ContextRecord->XIP);
        }

        // false tells it to continue as normal
        return false;
    };

from polyhook_2_0.

Top notch report, thanks!

from polyhook_2_0.