man.omnios.org/man9f/Intro.9f if you'd like a web link pending the one on illumos.org being fixed.

i found it (the VM I had handy also predates the changes, so I just pulled it up from the version in git :P)

Dumb question: Does overlay(7) actually belong in overlay(4) ?

it certainly seems it might belong there

does mdb use any metadata to figure out which modules to load, or is more or less based on the filename of the .so ?

[illumos-gate] 15635 /proc/pid xregs file panics cross-thread -- Robert Mustacchi <rm⊙fo>

danmcd, tsoome: No, because it's not documenting the driver.

It was trying to document the overall system.

Like smf(7).

jbk: I don't believe there's much metadata other than the SONAME.

ok

oh, I see.

So it'll walk the list of shared libraries or kernel modules and look for things with a matching name, IIRC.

I have two drivers that use (more or less) the same definitions, so using CTF, basically the same code can be used against either

though I still need to see how mdb handles resolving the types with the same name in the same image

I thought maybe module`type, but then I remembered I had to use ::typedef on my test case, so that is perhaps not a valid test :)

sommerfeld: were you fixing some of our printf/scanf disasters?

well, I nearly have a nightly working with *printf marked PRINTFLIKE

so, yes.

amazing, I won't start doing things like that then :)

currently about 380 files touched but I think some of them are unnecessary.

haven't touched *scanf yet.

gave up on a few things (anything using %ws or %wc, which gcc doesn't know about).

and tbl which assumes all real pointers are > 128 and stores characters in pointers.

I'm still hoping we can just use groff, long term, because fixing that stuff for aarch64 is at best going to go badly

rmustacc: Thank you.

(Man page header says "device" so that's why I asked.)

but I look forward to merging with you and discovering just how much crap I have to fix over the top of what you fixed, it should make life a lot easier for me.

real question is what granularity of review makes sense for all this.

then there are real atrocities uncovered by this (like the use of %wc in ftp's implementation of telnet options negotiations..)

we can teach gcc about %ws and %wc, going forward.

okay, my last nightly had a working -nd build, and then the DEBUG build tripped over a few things that were #ifdef DEBUG..