hrm...

so smatch seems to get upset with using the historic `uint8_t foo[1];` instead of `uint8_t foo[];` for a flexible array (at the end of a struct)

when you try to cast it to another type

annoyingly this is in sys/sata/sata_defs.h which is shipped... though i wonder if we just made a #ifdef _KERNEL version if that might avoid any binary compat issues (it really should be internal to sata.h, but...)

shipped doesn't necessarily make it not private

(I haven't checked, I'm just saying)

then how do you tell?

tea leaves

mostly :\

i mean.. i don't know if <sys/sata/sata_defs.h> was ever considered public by Sun, but if you look at it, it really probably should be considered private regardless of what it was (I can't imagine what use it'd be to anything outside of sata.c)

that's what I was thinking too, sometimes people shipped headers without really thinking about it, etc.

github code search or something might build confidence that nobody _actually_ uses stuff

i'm adding translation of several log pages, so specifically, it's the definition of 'struct log_parameter' in there

right, it seems like none of this would really be public, but I don't know.

i guess I can roll the dice on changing it and see if anyone speaks up

i mean, it's definition of scsi structs in a sata file...

it would be really nice if we could get those fixed.

Does illumos bhyve implementation uses any other type of virtual networking, but virtio ? There is Debian software, Openmediavault that can't see virtio networking and does not do hdcp resolving well post install, so i were thinking of another type of network interface in HVM, so that it can configure itself

You could try setting netif to e1000