CRASH post-processing a trace with many markers
I recorded a drcachesim trace with -record_function and a hacked drwrap where post-call hooks were sometimes missed. I hit a crash in post-processing where drwrap detected that a bunch of post-call callbacks had been missed and it caught up, resulting in a long series of function tracing markers. raw2trace has a 64-entry array for each insruction, under the invariant that one instruction couldn't possibly have that many entries. It gets overflowed, clobbering subsequent memory.
I think the marker loop in handle_kernel_interrupt_and_markers() should have its own output, to solve this and make the code more robust.