Untitled

[Naville]

Hey:
Regarding the Post LLVM9 support. You'll just have to edit how you initialize Legacy PMB.
Example:
Remove Hikari's Obfuscation.cpp, register passes as-is into LegacyPassManagerBuilder and instead of createSwitchPass() and run it, just add lowerswitch pass before the pass calling it into the pipeline.

For example, the old method was MPM.add(createObfuscationPass()), instead of that , you should write MPM.add(createLowerSwitchPass()); MPM.add(createFlatteningPass())

[Naville]

Also I'm more than willing to open-up permissions with the master for you folks to operate on it instead of this folk

[Naville]

From personal experience, injecting into Apple Clang sometimes results in other weird crashes related to uninitialized Apple Clang internal stuff (IIRC), not sure if this has been fixed in your build. -mllvm -load should works fine though

[woachk]

Hello,

We currently register the LegacyLowerSwitch pass from inside Flattening.cpp at

HikariCore/src/Passes/Flattening.cpp

Line 51 in f4307d2

FunctionPass *lower = createLegacyLowerSwitchPass();

We did split Obfuscation.cpp as two separate parts, PassScheduler.cpp and two different ways to register, a Loader_Hook one for Apple Clang and a Register_Passes.cpp one for outside Apple Clang. We didn't have issues with that and Apple Clang for now. Both ways allow to just register with -Xclang -load, with no changes needed to the host Clang at all.

About removing the PassScheduler, it means that we would have to define the arguments to enable/disable a pass inside the passes themselves right?

Thanks,

[Naville]

We did split Obfuscation.cpp as two separate parts, PassScheduler.cpp and two different ways to register, a Loader_Hook one for Apple Clang and a Register_Passes.cpp one for outside Apple Clang. We didn't have issues with that and Apple Clang for now. Both ways allow to just register with -Xclang -load, with no changes needed to the host Clang at all.

Should be good enough. I personally hate duplicating LowerSwitchPass which already exists in LLVM upstream, although that's probably just me.

About removing the PassScheduler, it means that we would have to define the arguments to enable/disable a pass inside the passes themselves right?

Yes, or put them in PassManagerBuilder

[nullpixel]

@Naville: hi, love your work on hikari. i’d be happy to port over our changes back into master on your original repo if you’re fine with that! do you have anywhere we can reach you to discuss things better?

[Naville]

@nullpixel Yes pwn20 should have my Discord.

[nullpixel]

just sent you a friend request!

[Naville]

@woachk Hey I've sent out invitation for master access. You can also use this link https://github.com/HikariObfuscator if you didn't receive it