I personally like the -A instead of just -4/-6 as indeed there are more adres families than just AF_INET and AF_INET6

but yeah I guess it's personal preference and muscle memory

but I do agree consistently is impotant

[illumos-gate] 14815 nvmeadm: identify support for namespace management -- Hans Rosenfeld <rosenfeld⊙gho>

can someone confirm a quirk I only just noticed, and I can't tell if I'm just losing it, or if I should have known this for literal decades and never did.

does /proc/self show in `ls /proc` for y'all?

It does not

I filed #14962 (fenix?)

BUG 14962: readdir(3C) of /proc should list /proc/self (New)

↳ illumos.org/issues/14962

I wonder if it's a weird compat thing, and I've just never known

(it seemed best to figure it out in a bug report so we have a record)

I wondered if it was to do with how closefrom() was implemented, but it looks like that has a check in it to avoid non-digit things there

yeah, my comment in the bug is maybe it was so you could list pids and not be careful that way

but it is possible that this is deliberate so that everything looks numeric with readdir()

but that seems so unlikely

might have to light the alanc signal

I wonder what would happen if we did what BUGS in proc(5) says we're going to, too

and kill old-proc

not that I want to, just that I wonder who still uses that.

what about inode values?

(granted it could probably show the value of whatever /proc/$$ shows)

it's a symlink, and it doesn't conflict because it's inode is 1

though there are other weirdnesses in /proc with readdir

we tried killing old-proc in Solaris 12, had to partially undo when we made 12 into 11.4 because there were still too many users

richlowe: now I'm worried the alancsignal is real

who said it isn't?

I see raf added /proc/self in PSARC/1996/463 but didn't mention not listing it in readdir - but I also don't see code in pr_readdir_procdir to specifically hide it, just none there to include it, so maybe it was an accident that worked out to keep the output nice for passing to the various proc tools, kill, ps, etc.

unfortunately, between Solaris 2.6 & 11.4 we failed at actively deprecating the old /proc interfaces, and left the default compilation mode to use them - only people who read the man page and saw there was a new interface available and changed their code to use it got the new stuff

too much backwards compatibility

does illumos/solaris have something similar to 'time -v' (linux) or 'time -l' bsd? I'm specifically interested in the peak RSS of a command

I don't think so, I only know of ptime -m which gives microstates (but not that)

sfe.opencsw.org/node/132 shows that we didn't even get other Sun projects to move off old-procfs

and it looks like tomww didn't upstream that fix yet: github.com/LibreOffice/core/blob/ma…sal/osl/unx/process.cxx#L1019-L1056

openzfs/zfs #13861 hmm, wonder if we're affected by this

should be easy to look and compare? I'd guess probably

jbk might know more

on a completely other topic, gitlab.freedesktop.org/xorg/xserver…086d02b80549981d205fb1f495edc373538 is an example of code in the wild that needs rmustacc's recent asprintf fix (issue 14933)

richlowe: i don't, but a quick look suggests we are