GDB single stepping allows threads to bypass svcSendSyncRequest

[misson20000]

If a thread is waiting on svcSendSyncRequest, it can still be stepped using GDB. It should not be allowed to be stepped forward until a reply is received.

@roblabla

[roblabla]

You're right. I should either:

• check if the thread is active before allowing it to step (that'd be the best, as it'd allow checking the registers for a thread) and return an error if it can't step (or continue another thread). This will make the flow a bit more complicated.

• Or I could just make it impossible to switch to an inactive thread.

[roblabla]

Actually the problem lies deeper than that. Basically when we trap the processor on a syscall, instead of replaying the instruction (putting PC back where it was), we go past it. For instance, if we have only a single thread stuck in svcWaitProcessWideKeyAtomic, we can CTRL+C to break the processor, and then continue to bypass the syscall.