i'll try adding that to the makefile (LD_DEBUG=libs ... i always forget about that) and see if that helps

out of curiousity, has anyone messed with building illumos-gate with libressl instead of openssl

elwisp: ping, do you use oxide computer? in the chat log, I read the message that you are using helios, it is oxide’s os for oxide cloud computer, I’m interesting the default raid for the pool, would you please show me the ‘zpool list -v’ for the node? a big thanks

tozhu: its just Helios in a VM that i am testing

tozhu: ill sent it to you in PM

jclulow it will cause recursive mutex panic

jbk I am pretty sure that racktop do - illumoe 13909 (fenix)

illumos 13909 fenix

BUG 13909: Want support for building against LibreSSL 3.X (In Progress)

↳ illumos.org/issues/13909 | code.illumos.org/c/illumos-gate/+/1581

andyf: actually all of that was torn out because of recent openssl changes

(which is incredibly disappointing)

Out of gate? I missed that happening.

(if so)

whatever is there apparently has issues with the latest bits in ways that weren't explained

tsoome: i know i've asked this before, but before i forget, is there any sort of upstream for efiapi.h ? I have a branch with a bunch of newer EFI GUIDs (mostly just so you can easily identify what protocols are present on a system)

which I added by hand, but if there's a better source, I'd like to use that

(github.com/illumos/illumos-gate/com…..jasonbking:illumos-gate:efi-guids if you want to see)

rmustacc: a while back we had talked about how to represent an internal M.2 slot for an NVMe drive in fmd/fmtopo... would the changes you recently integrated include bits to better express that?

(my recollection is that you thought you had some stuff you were working that'd help... wondering if this was it)

jbk - I have done work to add support for openssl 1.1, and then for 3.0 but I have never removed any libressl guards or related. When I did the work for 1.1 I regression tested with 0.98, 1.0 and 1.1, but not libre so I expect that patch in 13909 is still needed. If anything was torn out I am not aware of it.

andyf: i meant in our stuff in our build, not upstream...

Ah, ok, that's a relief.

sorry for the confusion

still trying to wake up after a marathon day yesterday

($@#$#@$@# perl)

jbk: I certainly have a better idea of how I would represent the M.2s and can point you to how we did for our system, with caveats.

that'd work

The biggest set of caveats is that we're trying to not put everyhting in hc and are creating different trees for that.

So hc strictly represents mechanical releationships, devices in slots, bays, sockets, etc.

However, we've also separately added things like the a pcie tree and more to try to have something that is representable (not just on our system, but more generally).

I can shoot you an e-mail with more details?

yeah that's fine

i'm kinda hobbled right now because my comcast has been down since late last night

so my laptop is tethered to my phone for work

but means i have to disconnect to reach my home network :)

so it'll be later before I can dig in

(my goal is i want to represent the stuff in my home server)

My ISP sent me an email today saying if I want to keep my static IP, I need to click on a link in the email to let them know. Clicking on the link immediately changed my IP address at the ISP side, apparently so I'm still recovering.

I /do/ have IPv6 now though, so that's something

i have the front drive bays and USB ports where they're identified now with what you'd see on any product documentation or physical labels on the system

but the network ports and the M.2 slot were the last bits

about 2 months ago, my internet down for like 10-12 days

it'd come back for like 45 mins, then die for hours

I just re-discovered fire using the incremental file in nightly, my a8 take like 5 hours to build-gate now just 2 using incremental

andyf: welcome to the new world with more odd network quirks now that you have IPv6

once things settle

i should dig up some issues 😅

Well, I live in a mostly IPv6 world at work, so about time I caught up at home :)

andyf we need an additional reviewer for the mdb vmm change so dan could approve it for RTI, I don't know who else could take a look

neirac - it looks like Jordan added herself as reviewer a while back, I can ask her if she still intends to review.

andyf thanks a lot.

I have one question regarding this doc illumos.org/docs/contributing/gerrit in section 2, to copy the hooks I always need to add -O flag to scp if I don't do it it does not work, I'm in omnios. Someone else had that problem?

you copy that once... well, unless you have multiple repositories

tsoome oh, yes but when I first setup this on a new machine and copy the hooks scp does not work without the -O flag, is that flag required?

what does -O do? man scp does not seem to mention it....

ah, other system does... Forcing the use of the SCP protocol may be necessary for servers that do not implement SFTP

so it depends on server

yes, for me the -O flag works on code.illumos.org

should be just scp according to the documentation

or my install is the problem, just asking if someone else needs to use -O

is not bad, just that I forgot about it as I only needed once, but when setting up other machine I bump into it.

Is it required to do a non-debug build to create a onu?, or I just could skip the non-debug build?

depending on change, you may want to test both (thats the onu is for)

But strictly speaking, multi-proto is never required to use onu.

You either pass in debug or non-debug packages.

During early development, for example, I will pretty much only ever build and use debug until the end. Some people are the opposite.

rmustacc , tsoome_ thanks, I was building debug and non-debug for a change and taking a lot, I'll work with debug only then.

Depending on what you're doing, I also find it helpful to just use dmake install in the single package I'm working on and then rebuild packages.

*single directory

rmustacc oh yes, on changes on mdb I do that, I was finishing adding a tcp socket option so I assume needed to do a build

It's always safer.

out of curiousity, any thought given to the 'next' version of gcc to target for building?

you mean, for gate build?

yeah

13

IIUC, 13 actually support plugins -- I wonder how interesting it'd be to do our changes (e.g. -msave-args) as plugins

There's a PR up on github/illumos/gcc that I need to get back to. It's just carrying forward the patches though at this point.

I have some 22 or so cleanup patches still in queue with gcc 12, 13 will add about 10 more, with *current* warning blockers

but omnios has gcc13 and userland is built with it these days

13 because of C23

well, it will enable us to walk that path;)

anyhow, time to get some sleep...

just saw the first round of patches to clean up gcc 14 warnings in various open source projects go by - apparently -Walloc-size in gcc 14 pays attention to the order of calloc args

i have a change that cleans up most of the warnings in mac... there's a small number i'm unsure of, so those are in a separate commit, but I should probably get those up once I've had a chance to run them through some smoke tests (aside from some missing cases in a few switch statements, mostly just assigning default values)

smatch does that too, but I be suppress that in some places at least.

*I bet we

jasonbking/illumos-gate a837689 for the curious

is there like a rosetta stone to compare our tcp struct field names vs the one in Linux?, I thought of trying to search comments in the lx brand commits

I wouldn't expect there to be.

Is there something you're trying to figure out how to implement?

rmustacc yes, I'm missing to map the values from tcp_info, I'm implementing TCP_INFO socket option I'm missing just returning the right values, here smartos.org/bugview/OS-4525 is the tcp_info struct but the names don't match or I'm missing to find the ones needed in tcp_t

→ OS-4525: lxbrand implement getsockopt(TCP_INFO) (Open)

right now I'm just returning made up values, but not ideal

I see. That's not going to be a straightforward question to answer.

There's a lot of research required.

rmustacc yes, I was reading how are those calculated, and maybe some fields need to be added to get the information, I'll staart with just the fields that some applications require

Are there applications outside of lx that are looking for it?

rmustacc not yet

rmustacc I'll try to read a little more about the fields in other platforms and how they map those values.