I will try and have a poke at it tonight

thanks!

sommerfeld: It may be alright now (touch wood)

[illumos-gate] 17612 pmap -A does not work on core files -- Jonathan Perkin <jonathan⊙pou>

Yep, seems to be better.

Had one "error: remote unpack failed: error Missing tree 0869e8046850634dd5e0031fe608f855e35c973c" which cleared up after a "git fetch --all".

Thank you!

Thanks!

looks like `git whatchanged` has been deprecated in git 2.51 or earlier; documentation says it's just syntactic sugar for `git log --raw --no-merges`.

and the "--no-merges" part is probably not what we want (nobody should be committing merges to illumos-gate/master)

see illumos/docs #103

I think we could probably say in the face of a gerrit URL it's not required.

`git show --stat`

Hmm. `git show --stat` doesn't include file modes and hashes. `git show --raw` output looks like `git whatchanged` for a single commit.

(my professional paranoid wants to see the hashes there...)

I think the hashes are just noise, it's not like we're inspecting them

Ultimately we expect people to put things on Gerrit, so, a link to a gerrit change is totally fine instead of pasting that stuff

I'll update the pull request in an hour or two based on the discussion here (assuming it converges)

[illumos-gate] 17645 add trousers/32 and libdbus/32 to packaging.deps for builds on openindiana -- Bill Sommerfeld <sommerfeld⊙ho>

sommerfeld: Despite what I think sometimes people suspect, we're definitely keen to remove an unecessary faffing about

*any

Some of the faffery is necessary but yeah, if you've got a gerrit link, and that's where we're going to pull in the change from, it seems better for us to just click through and look there

also attaching patches not really required etc, in that case

It's pretty easy to poke at all the stuff in gerrit, e.g., via illumos.org/docs/contributing/gerri…errit-using-more-advanced-git-tools

I'm slowly coming around to that, esp. given how the "cherry-pick" download-patch method works really well for, "Let's build -gate with this."

I used to have, but deleted because I'm a fool, a refspec that pulled all the gerrit reviews into references

I think jclulow told me how

all I remember is you have to tweak the "fetch" on the remote

Hi all! I recently posted sth to illumos-announce. Is there something special I need to include, or am I just waiting for moderation?

but I had basically everyone useful or interesting as a remote, and then cron fetches

vab: it'll be moderation

richlowe tx will be patient then :-)

vab: might be better, rather than patience, to ping someone who moderates it :)

sommerfeld: thanks for digging out MICRO SIGN being the wrong micro sign, by the way

richlowe Yeah right. Topicbox says announce+admins⊙lio -- I see that there hasn't been any traffic on -announce for more than four years... I guess I'll just use -discuss.

[illumos-gate] 17615 du: using -A should still output blocks, not bytes -- Toomas Soome <tsoome⊙mc>

Thanks for the update @vab !

danmcd np :-)

BTW we since have found that a Sun SPARC T5-2 will still crash booting OpenIndiana if it has two CPUs.

Looking at systems that work, the current working hypothesis is that there is a limit of 192 CPU threads.

More than that -> OBP throws "Fast Data Access MMU Miss"

that has definitely worked before, on the bigger T3

it's not an NCPU type situation

um... remind me, at what state you get it?

s/state/stage/

tsoome_ Just after it prints the "Ethernet address" line during boot.

after obp banner?

richlowe Yes, it's strange. A T5440 works with three CPUs = 192 threads but not with four = 256 threads.

so it smells like error is thrown while in boot block.

tsoome_ Yes, it gets as far as printing "OpenIndiana Hipster 2025.10 ..."

ah. thats quite a bit later then.

If people are interested I could ask olbohlen to post a boot log somewhere. There's not much to see though.

I have updated code.illumos.org/c/illumos-gate/+/4398 with the small letter mu

→ CODE REVIEW 4398: 17642 illumos sources should be consistently utf-8 (NEW) | illumos.org/issues/17642

well, tbh, I'd still put my time in arm;) but IMO there is three basic options - debugger, debug printouts and any available source about related platform to compare with. unfortunately "Fast Data Access MMU Miss" can happen on different things, so it too obvious error to look after.

in the closed-bins, you'll find forthdebug, which is basically a debugger of last resort

while they're closed bins, and obviously you should be a righteous person, iirc they were mostly just text inside and if they needed updated to be used that could be something you did for yourself and didn't redistribute or admit to

richlowe: yeah I had something like that refspec when I was using gerrit at Google. I believe you want to add something like ` fetch = +refs/changes/*:refs/remotes/illumos/changes/* `to the appropriate [remote] section of .git/config

you can have multiple `fetch = ....` settings per remote

richlowe: oddly it looks like the unicode docs are inconsistent on whether to call it GREEK SMALL LETTER MU (found in unicode.org/charts) or GREEK LETTER SMALL MU (found in unicode.org/reports/tr25)

i never knew that there was one preferred over the other one. one used as a unit the other one as the greek letter. this issue is more interesting than i thought initially

well that's about U+03BC v. U+00B5

richlowe We're all outstanding and righteous community members. :-) In fact there are two tests planned:

1. try to build a debug kernel

2. Set up a T5-2 with S11.4 CBE as control domain and an OI LDOM, then watch what happens as more threads are allocated to the LDOM.

hadfl: my experience looking into this confirmed my prior suspicion that unicode is a big inconsistent hairball.

the various spaces are fun as well