can anyone think of any reason why mdb_ctf_member_info() doesn't use ctf_member_info() and instead does it's own iteration?

ping @cross

NVM.

does bhyve support RDMA/RoCE in illumos/smartos ?

jbk: it appears it's doing its own thing so it can try to manually resolve forwards.

mdb_ctf_type_resolve() does more than ctf_type_resolve() at first blush.

Maybe a dumb question: can gcc12 on illumos build binaries for a linux x86_64 target?

if you build a cross compiler targeting linux x86_64, you can certainly use it to build binaries for linux x86_64

Woodstock (LIBERA-IRC): Thanks, I was hoping such a cross compiler could already be packaged in e.g. pkgsrc

I see cross-* for different CPU archs only

in theory sure, in practise it's unlikely to be worth the effort given the amount of time it takes to build and the likelyhood of folks wanting different targets

a set of pre-built targets available somewhere might be useful though

jperkin (LIBERA-IRC): I thought so, thanks

The AVL tree code's TREE_TMP() is a really clever way of getting -1, 0 or 1 to indicate how two numbers compare. Unfortunately, a simple subtraction that gives a negative value, 0 or a positive value is always faster. :/

s/TREE_TMP/TREE_CMP/

Now I see why it is done this way... the subtraction trick doesn't work with unsigned integers.

ryao: it can also fail if the diff is too large

Good point.

Thanks for the help, everyone!