toasterson - it might not be the way you want to go, but there is an illumos branch of cloud-init that we use in the omnios cloud images. Best way to be bug compatible with cloud-init, and it works well in AWS, Azure, DO, bhyve, ...

andyf: Are the dependencies still borked? It has this one unmaintained C dependency which I just could not get compiling on OI. I don't want to go the bug compatible route anyway. I am interested in Network config and getting something that can bootstrap the native config process. And metadata agent does that. I wanted to extend it for some time but since I can now drop in something decent network config it's enough to scratch my itch. It's

not a cloud image but it's enough automation so that I dn't have to run the installer manually.

I don't remember dependency problems - we've been shipping it in omnios for a few releases now

Yeah for some reason omnios doesn't have them

I can try again and see if I still hit it though

21.x was the last I tried

cloud-init was not too hard to port, and then we get all of the benefits and the different cloud provider support all for free

Well the porting part no, that dependency however is so deep in the code I'll have to rewrite how cloud-init finds network interfaces to get it out

I wonder why we don't see that on omnios. Interesting

IIRC that one dependency got a non version bump change last I checked and you have a package from before that.

I might need to pkgrecv your version from the repo.

and steal that

Well, in anycase. Splitting the user creating parts off into its own section worked. Now to do some testing

jclulow: richlowe Have the two of you any opposition against me refactoring image-builder into a library binary hibrid package? I need most of the functions in there for github.com/OpenFlowLabs/aurora-open…cloud/blob/main/imgbuild/Cargo.toml and poaching/copying them seems a bit a waste.

[illumos-gate] 15290 lib: remove pragma ident -- Toomas Soome <tsoome⊙mc>

ding:)

toasterson: I can think about how we might do it? I'm wary of creating a committed interface to be honest, especially without doing some of the refactoring I'd already like to do.

Well I as long as it's just us we wont need a commited interface. But I'll gladly follow along with the guidelines. I have a version for now but cant upgrade now as image-builder and my refactor have diverged so heavily I'll have to redo everything anyway.

Sould we start a IPD to flesh out what Patterns we want in the libraries?

Side note to channel: Somebody with permissions should delete illumos/illumos-gate #87 as SPAM

The spam thing is on my list yeah

toasterson: I don't think we need an IPD. It would help if you could make a brief list of what of the existing behaviour you would want exposed as a library. Mail or gist would be fine

I'm also keen to replace the JSON with a KDL document for instance

jclulow: Then we can use my imgbuild work in parts

I focused on zone builds but it is KDL based

Do you have an example document I could look at

I just rebased code.illumos.org/c/illumos-gate/+/2619 past 15290 (would have RTI'ed this last week but getting set up to test it took longer than I wanted).

→ CODE REVIEW 2619: 15356 closefrom(INT_MAX) misbehaves and closes all descriptors (NEW) | illumos.org/issues/15356

now, where did we stop with cw...

sommerfeld: Are you looking for review sill?

actually, looks like gerrit didn't invalidate the +1's after rebase. thought it would.

I don't think I need more eyes on it then.

No, if you just rebase and don't change anything (commit message included), it won't invalidate it.

the diff changed but the final file didn't and I guess the latter is what counts.

(both 15290 and the previously reviewed patchset of 15356 deleted the #ident from closefrom.c)

It looks like it didn't detect a change from patchset 2 to 3, which is I guess what it'd use.

Ah, interesting. Well, that's lucky I guess.

yup.

Kind of surprising.

must be comparing the git blob hash for each file instead of diffing the diffs.

jclulow: github.com/OpenFlowLabs/aurora-open…czone/testdata/zones-base/build.kdl

Several infact the whole testdata folder contains samples

Huh, thats an old example....

jclulow: github.com/OpenFlowLabs/aurora-open…ain/opczone/testdata/image_base.kdl is more current

[illumos-gate] 15356 closefrom(INT_MAX) misbehaves and closes all descriptors -- Bill Sommerfeld <sommerfeld⊙ame>

[illumos-gate] 15239 sharing pcieb HP/LBW interrupts with AER results in spurious ereports -- Keith M Wesolowski <wesolows⊙oc>

[illumos-gate] 15269 pcieadm misrepresents PCIE_AER_CTL as PCIE_AER_RE_CMD -- Keith M Wesolowski <wesolows⊙oc>