Lots of people use those on all kinds of systems, so, I would be pretty surprised

right

it is not trivial to create an EFI disk label.

Usually one is created for you automatically when you use "zpool create -B"

Though you can also do all those things manually with format

I could not wingle format into doing what I wanted

zpool create -B isn't what I want to do. I want to create a separate partition for paging. I think I will at the very least label the disk on another OS.

It is your computer!

I know paging partitions are an antipattern on illumos, but I no longer have trust in the patterns that illumos OSes prefer regarding this.

I'm well aware of the disk caching ... thing, that motivates feeding whole mass media to ZFS.

[illumos-gate] 15121 tst.temporal.ksh has PATH dependencies -- Robert Mustacchi <rm⊙fo>

[illumos-gate] 15126 ipv4info_t translator produces bad flags value -- Robert Mustacchi <rm⊙fo>

[illumos-gate] 15110 mdb ::printf replicated mdb ::print bitfield mistakes -- Robert Mustacchi <rm⊙fo>

my keyboard died.

I think it's back together.

ultimately the success came with a USB CDROM.

[illumos-gate] 15109 dtrace replicated mdb's bitfield mistakes -- Robert Mustacchi <rm⊙fo>

tsoome_: I filed a follow-up ticket for updating boot(8), are you otherwise happy with code.illumos.org/c/illumos-gate/+/2498 ?

→ CODE REVIEW 2498: 15137 ZFS should allow direct import of a root pool from a /devices path 15122 vdev_disk_preroot_fini can race against the child vdev open taskq (NEW) | illumos.org/issues/15137

beep boop (:

I will definitely look at the crash dump, sorry, it's just an incredibly busy week at Oxide.

That's ok. How's your day been thus far?

or, too busy for smalltalk?

Hectic!

jclulow +1 wit h small nit:)

What! I love composite expressions haha

jclulow I'm cleaning tons of such constructs with clang (because of missing compare operator)

:D

thats the "using the result of an assignment as a condition without parentheses" warning

Yeah I mean you obviously need the parens and the comparator

anyhow, not a blocker for me, hence +1

Tah

btw, if you have a bit of time, this one might interest you: code.illumos.org/c/illumos-gate/+/2314

→ CODE REVIEW 2314: 14919 tem: implement xenl (NEW) | illumos.org/issues/14919

i feel like i have been photocopied

hrm?

overexposed to light?:)

need to set up some review requests...

tsoome_: Have you tried vttest with tem

vttest? does not ring the bell

I once did run few tests coming with xterm

yes, I found it:)

seems interesting

out of interest did anyone figure out a way to run the full vttest suite without having to go through all the silly prompts?

I think you kind of have to use the prompts

I don't think it works without you visually inspecting the results

much of it would probably require visual inspection

Yeah

fair enough

heh:D you were faster on typing:P

Now I want double width and height lines haha

I feel like that is not commonly implemented

"fat" lines?

can't recall what the issue was but was trying to figure out if there was a problem with my terminal output and found it really frustrating to have to navigate the vttest menus instead of just running something once and then scrolling back in one go

jperkin: Yeah it definitely has a very particular style

especially when you aren't entirely sure what the problem might be and which test might trigger it

tsoome_: Yeah, like invisible-island.net/vttest/images/…TTEST-double-sized%20characters.png

ah, stretched.

oh yeh, it was recent gnupg or something related breaking output in pinentry-tty mode

that would be more work:)

the current code is assuming fixed size glyphs:)

actually while I have some tty experts handy, do either of you know what might cause this? mail-index.netbsd.org/pkgsrc-users/2022/08/31/msg036134.html

jperkin locally there means what?

on illumos console?

running direcly on my mac in terminal.app

usually such mixup hints about bad utf8/unixode sequences

unicode*

interesting thing is that vertical lines are ok

yeh that's what I found odd

and the corners..

and I'm almost certain it used to work fine, ncurses didn't change during that period (so the terminfo entries should be the same), so trying to figure out what code changes may have affected it

but other than forcing different TERM I don't know where to start looking with this stuff

hence hoping vttest would print "TERM is missing the frobnicator capability" or something

It would be interesting to know if it were using Unicode box drawing or whether it was using the VT style mode switch to b

to box drawing alphabet and back

'g' is 0x67 and '>' is 0x3e, so you should see some byte sequence ending with those values - while drawing horizontal lines, that is

for unicode, you can test with ficl-sys (in illumos) like this: 0x2500 xemit cr

this will draw horizontal line

in shell... with octals like this: echo "\0342\0225\0220"

thats with utf-8 encoding of U+2550

You may want to check what locale environment variables the Mac is passing through SSH

hello, guys, please look at the man page: illumos.org/man/3C/pthread_rwlock_destroy

int pthread_rwlock_destroy(pthread_rwlock_t **rwlock);

the argument should be pthread_rwlock_t *

Can you file a bug please

yes, of course

Thanks!

sigh. we have long way to make vttest ok;)

and vt52 tests did crash us. not good at all.

→ BUG 15169: incorrect argumnet declaration for pthread_rwlock_destroy (New)

Thanks denk

tsoome_: crash, like a panic?

it was kind of instant, but this UTM vm is able to boot like at 5 times from 10, it is a bit effort to see if we got dump...

I havent had any time to find out why it will hung during the boot...

is UTM basically QEMU?

yes

with machine emulation - emulating x86 on arm

Ah

I feel like QEMU has gotten increasingly screwy since about 6.X

*6.0

so that is probably contributing factor

but also, can not exclude bugs from our side:D

should I, one day, write my own imapd

Yes

(¿)

especially because freebsd on UTM does "feel" snappier, but also it does not hung during boot....

Does UTM have any debugging facilities? Can it inject a serial break or an NMI?

I would hope but havent really investigated

Once I get clang cleanup done, I'll have a bit more time for investigating:D

this loader serial port dropping characters thing kills me haha

so hard to type

that is bug in firmware:(

It feels like other boot loaders do not have this problem though

when serial port is bound to ConsIn, the SimpleTextInput is "stealing" from SerialIO

yes, it depends on how the console bits are done.

Can we do our console bits some other way?

what is your current console variable?

et value of*

err

err*

It is going to take me minutes of pushing buttons to find out haha

clang cleanup?

LXGHTNXNG to be able to build illumos gate with clang

is this some kind of subbranch

only in my tree atm

ok

is your tree published anywhere

and it is not particularly useful atm:D

tsoome_: console=ttya and os_console=ttya

I have to press each key between 1 and ~50 times (a different number of times for each key)

I had that in loader, in qemu (on virsh) after what I think was a failed attempt to fire up the OS

It happens in almost every virtual machine I have available and on some hardware

which was frustrating enough that I reset the machine

LXGHTNXNG github.com/tsoome/illumos-gate/tree/clang

right on

LXGHTNXNG just note this is work in progress

right

if there's anything I can do to help, let me know

jclulow righ, I was confusing myself earlier -- its not about serialio issue, but about the fact we are kind of forced to use ioport based serial port access (comconsole.c) from bios version. And this is because uefi does not provide good way to identify serial ports. Altho, I'm not 100% sure if this issue was biting with serialio too... need to test it again.

This is definitely a UEFI VM, we don't currently have BIOS/CSM for Propolis generally

The problem is, we do get serialio handles from handle lookup, but we do not have good generic way to build the mapping to tty[abcd]

I guess I should re-test if serialio will miss the input too and figure out the way to create mapping for tty?

this is also needed to support those memory mapped/pci/usb serial devices...

Presumably on i86pc you can look at ACPI tables

and if those are not present or don't seem correct, just fall back to the classic PC values for COM1-4?

It is not really clear if UEFI's vision of the serial ports is really helping

What is odd is that if you leave a bhyve VM sitting at the ok prompt on serial and forget about it for some time, it can work perfectly when you come back.

huh

That is odd

and when I did look at this, adding small sleeps to bits of loader made it work properly, but I didn't get further

the grub2 is just creating "serial" device and does not really concern itself about mapping, freebsd is other way complicated

fbsd is basically using the same "solution" as we do - using comconsole.c from bios version on x86

jclulow is your case from VM or plain hw?

this one right here is a VM

but the serial stuff works fine once you get to kmdb

using edk2 based firmware I guess

It is ... OVMF?

yes, with kmdb we have exited UEFI boot services and own the hw

aye. well, in core, the root cause is that efi simple text input protocol is "stealing" the serial io input, so that there is a race with INB instructions from comsonsole.c

So does this happen on FreeBSD as well under similar VMs?

yes, but they try to enforce you to use console type "efi" if you have serial port multiplexed in ConIn

Can I "set console=efi" on our loader?

no, we do not have "efi" console

ah ok

in fbsd, it will fall back to use SimpleTextOutput protocol, but that one is other kind of evil

haha

It would be nice to somehow just kick the UEFI input handling in the shins

SimpleTextOutput will send to all ConOut devices, but it will not pass esc char, so you can not use CSI

I will try to figure something. Obviously this has to get improved

I stand ready to test!

need to get my wife to shopping now:D

That seems prudent

weird, a fully loaded VM will only seem to use 20% of my CPU

ah, nvm

it's not fully loaded at all

[illumos-gate] 15137 ZFS should allow direct import of a root pool from a /devices path -- Joshua M. Clulow <jmc⊙oc>

sry qrong #