more tidbits and context

sync & add a failing test

sync & commit

while there, clang-format the code I touched

rebase + try to use this diff from a different machine

I can't see the file anywhere, so someone must have deleted it.

Most of my comments are minor enough I'd be OK if this went in. But please consider them before committing.

Sorry for the late review, and for destroying your diff.

I would probably add more tests for the different configurations, but I suspect more diffs are coming after this.

sync again?

sync again - something's wrong with Git

sync to HEAD

add errno printout, in case something goes wrong on the buildbots

sync

std::unordered_map -> llvm::DenseMap (as requested)

sync

reorder like it used to be

nth_element doesn't actually order things on either side; put them back

sync

sync

FWIW, I've been seeing signal handler traces written between ENTER_ARG and ARG_PAYLOAD.

Although it looks better than the original, there's a lot going on and I can't guarantee I haven't missed some stupid typo. Basically, LGTM.

still not 100% finished - I'll get to it tomorrow. I am convinced the overall concept will work though.

partial review only

LGTM now. Let's see what breaks :-)

I still don't like how we handle TSC vs. emulation here, but that's not the point of this diff.

break up RUN: lines in the test

Technically, the pthread functions can return EAGAIN or ENOMEM but it (ehm) "should not happen in this case". Really. I mean, come on!

I'm wondering how much will all the malloc(3) calls in one thread impact the thread locality work done in tcmalloc. Better ask a tcmalloc expert though, I've lost track of what's going on.

latest update before commit

address comment

address yesterday's discussion

rebase to HEAD

rebase again; tests now pass

I like how we're moving forward to C. LGTM.

rebase to HEAD

I will wait for the other diff to go in then. In the meantime I can sort out my new system's toolchain issues.

Nitpick: maybe update docs/XRayFDRFormat.rst in the same diff.

I don't understand why do we need the version number bumped but I'm guessing you'll show me a counter-example in response to this comment.