Enumerize tracepoint categories as described in #314

[syogaraj]

1. Used trace-cmd list -e for the list of tracepoint categories.
2. Tracepoint now has its own file (tracepoint.rs). Hence, removed TracepointOpts from program.rs and placed it in tracepoint.rs.

[danielocfb]

Nice! Would you mind splitting this into two commits, first the introduction of the tracepoint module and the move of TracepointOpts and then the addition of TracepointCategory? Also, please take a look at my comment for the Unknown variant.

[chenhengqi]

Looks good overall. But first commit is incomplete, you can't not refer to TracepointCategory::Syscalls defined in the second commit.

[chenhengqi]

A second thought. Could we just add some generic categories ? No arch-specific, No driver-specific.
See https://github.com/torvalds/linux/tree/master/include/trace/events

[danielocfb]

You are saying remove everything that is arch or driver specific? I have no strong preference, but to me that seems like a moving target: over time something may become generic. It also seems somewhat arbitrary and ultimately we just provide possible values, but runtime detection by the kernel should have the last say anyway. So personally I guess I'd err on the side of making more values available.

[chenhengqi]

This list seems taken from some random running machine, especially the amdcpu thing.

There are tens of arch and thousands of drivers, trying to keep them here also a moving target.

[danielocfb]

But first commit is incomplete, you can't not refer to TracepointCategory::Syscalls defined in the second commit.

Yeah, that's a good catch. Can you please fix that up and make sure everything builds at each point, @syogaraj ?

[danielocfb]

I see there are some conflicts with the merged #1115. I believe they should be fairly mechanical to resolve, though. Sorry about that, but we'd have conflicts either way.

[syogaraj]

But first commit is incomplete, you can't not refer to TracepointCategory::Syscalls defined in the second commit.

Yeah, that's a good catch. Can you please fix that up and make sure everything builds at each point, @syogaraj ?

Missed this @chenhengqi. fixed them now. @danielocfb

I see there are some conflicts with the merged #1115. I believe they should be fairly mechanical to resolve, though. Sorry about that, but we'd have conflicts either way.

@danielocfb, Resolved the merge conflict from #1115 and moved RawTracepointOpts to tracepoint module.