DTrace Hangout Notes 16 Feb 2017

Some OpenBSD folks are working on a non-CDDL version of the CTF tools.

James mentions that DWARF5 is out and Robert mentions that we might talk to the Linaro folks. gnn@ will talk to emaste@ and jhb@ about this as well.

gnn@ is working up a patch for the copyoutmbuf work. We will adopt the XNU #define for the Mac OS changes.

gnn@ asks whether the tests are now working and markj@ says that they are and that we should clean up the last few test failures

rwatson@ points out that issues on ARM32 cause the kernel to panic. markj@ says these are not surprising and that we'll have to look with the test suite.

gnn@ to talk to emaste@ about fast model licenses

gnn@ to ping ruslan@ about testing on RISCV

A question comes up about the blacklist of functions. This is done via strcmp() and there is a very small set, most related to locking.

James mentions that the easiest smoke test is 'dtrace -n fbt:::' as a way of catching functions that were not dtrace safe.

Domagoj continues to work on installing probes into guests on bhyve.

We discuss how we converge the file layout and we all decide to discuss this in an RFD. One question comes up about splitting out platform specific functions (#ifdef ILLUMOS etc.) into their own files.

A discussion comes up about whether or not llvm is mucking with DTrace's internal functions, such as memcpy().

James says he'll have to drop out of future calls but may be back from time to time.