typod 'embeded' as 'impeded', and I'm not even sure it was wrong.

jclulow, andyf: so yeah, my system, no seeds, no magic, with my bug fixed just starts importing manifests happy as anything.

not exactly pleasant, because time, so the seed is definitely good

I have suspicions I am just getting radically lucky, radically often.

which, you can imagine, was not at the top of my list of theories, for reasons.

a

[illumos-gate] 15139 loader: avoid duplicate load for environment and console-font -- Toomas Soome <tsoome⊙mc>

ugh, just got bit by this guy again: illumos.org/issues/15024

→ BUG 15024: NFS can exhaust pool threads getting RPCSEC_GSS credentials (In Progress) | code.illumos.org/c/illumos-gate/+/2402

all it takes is one misadventure with one kerberized NFS session to deadlock the whole system

oh there are other fun ways too :)

[illumos-gate] 15136 want unit test for bhyve dirty page tracking -- Patrick Mooney <pmooney⊙pc>

[illumos-gate] 15128 libnsl: variable is used uninitialized -- Toomas Soome <tsoome⊙mc>

can we just evict NFS forever?

NFS is my favourite

hot pursuit was best

ha

Underground

All time favorite car was from Grant Turismo PS1 - Mitsubishi GT3000 ... what a good time

weird, both gcc7 and gcc10 say "error: a label can only be part of a statement and a declaration is not a statement" about github.com/omniosorg/kayak/blob/ddb…f0a3eb8a4f53/src/mount_media.c#L330 despite -std=gnu99 being specified

that's talking about labels, not variable declarations

I'd presume somehow you're conflicting with the name of one

declaring the variable above the label (at the start of the function) gets rid of the error, but i wonder what's wrong with my gccs as apparently it works when building omnios

labels as in the things you goto to

I don't know, that's weird. I can see why it would seem odd to try to put a local on the stack in a label-dependent way, but it should get hoisted regardless

so I can see why the compiler would think that was odd, or possibly not ok, but I can't work out why it would choose to do so only for you

I'm looking at the code in 10.3 that does that, and I definitely can't see how it would depond on anything.

actually tried with gcc on omnios and clang on freebsd, all of them are not happy

not sure how it builds while building kayak in omnios :)

I would definitely not expect that to work

I was assuming it worked as a GNU extension, or by chance, since as I said, the variable would get hoisted anyway.

but looking at the code, definitely not.

I think there's some question of when the initialisation occurs

and whether it occurs again

yuripv: were I you, I'd assume it doesn't work on omnios either, and fix it.

yeah

declare the variable at the top of the block, initialize it at the label or whatever

I feel like case labels also have this issue

case labels you traditionally just introduce a block

doing that for other labels looks horrible

case label blocks make me grumpy because there is no good thing to do for the last arm

œ

Which I suppose is an artefact of our otherwise delightful choice to align the labels under the `switch`

C's block structure is kind of a convenient lie anyway

I think that's probably what we would put on C's tombstone

RIP C, a convenient lie

sometimes I wonder what it would be like to write a compiler that understood the peculiar Plan 9 dialect of C, on Unix.

anonymous members of structs. What?

I'm pretty sure GCC does

hm

Idk. Maybe I just have Plan 9 brain poison.

richlowe: yep, done, omniosorg/kayak #236

LXGHTNXNG: I suspect the early Go source history probably has compilers like that?

Also presumably 9port etc

I'm pretty sure p9p just relies on gcc having the extensions

richlowe: the Venti file server in 9port at least removes using Avl anonymously. Instead of that, though, it appears to rely on struct members being in the order specified (something the ?c compilers on Plan 9 do not guarantee)

It also relies on the C compiler not adding padding to the beginnings of structs.

You keep calling it a "C compiler" haha

I feel like it compiles Something Else though

It must be hard to have any kind of ABI if the compiler can just reorder all your fields

I suppose Plan 9 didn't care about that

or read blocks off data from a device

perhaps they had a `#pragma less_esoteric`

for that

I could see them doing it to minimize padding, and exempting structures with no padding or bitfields or something

everything is statically linked. Also I think I lied about the reordering

even if things are statically linked, they had all this "I read structs from files everything is a file" stuff

You have to imagine that some of that was done with a read(2) into a struct

[illumos-gate] 14978 ZFS autoexpand property should work for root pools -- Joshua M. Clulow <josh⊙so>

jclulow: they're very big on everything being strings, too

process exit stuff, signals...

That's possible. I vaguely remember looking at how some of the window drawing stuff worked and I had felt like there were some structs in there -- but perhaps I am misremembering