prasun3

While switching instrumentation, I need to ensure that the regs are restored. Therefore, I tried to change `clean_call` to use `DR_CLEANCALL_READS_APP_CONTEXT`. I ran into an assert `DynamoRIO usage error :...

Thanks! I have set short_reaches to false for now. Now I am running into cases where some GPRs appear to be clobbered. For example, the value in rdx here is...

Ok I see now that rdx is not restored when this jrcxz branch is taken. So ideally it should be restored after the jrcxz+jmp and before it gets to jbe?...

In the regular case, `drreg_event_bb_insert_late` calls `drreg_insert_restore_all` which restores both `rcx` and `rdx`. With `DR_CLEANCALL_READS_APP_CONTEXT`, `drreg_event_clean_call_insertion` has already restored `rdx`, so `drreg_event_bb_insert_late` skips this reg. All of this happens inside...

Looks like I was missing DR_CLEANCALL_MULTIPATH

I am seeing an assert in translate.c where it hits `ASSERT_NOT_REACHED` in `recreate_app_state_from_ilist` . This happens when we call `dr_flush_region_ex` . The application is a slightly modified version of threadsig...

It doesn't look like it is failing to translate a specific instruction. I think it is unable to find the mcontext->pc in the ilist. I have attached the log: [translate-failure.txt](https://github.com/DynamoRIO/dynamorio/files/8047784/translate-failure.txt)...

I was clearing `op_L0_filter` before calling `dr_flush_region_ex`. A fragment was getting re-translated on another thread while `dr_flush_region_ex` was in progress. Therefore the fragment length at `fcache_fragment_pclookup` was inconsistent with the...

I now have some basic runs start going through. I can start sending some PRs if the approach below looks fine. It is not perfect but we could add this...

> My main question here would be: should this feature use drbbdup for the two phases, rather than flushing? Yes I had considered that initially (https://github.com/DynamoRIO/dynamorio/issues/3995#issuecomment-825748234) and we had discussed...