eyedeekay
Most of you didn't have git-lfs permissions anyway but in case you did they're disabled now, because gitlab sucks
eyedeekay
If I had known I would end up running it I would have talked Mikal into using gitea
eyedeekay
Today is deadline day for any big changes
eyedeekay
Actually, let's give it until the 30th for major changes
dr|z3d
What are we talking about, eyedeekay? 30th of this month?
dr|z3d
And how does that align with the segmented db stuff?
eyedeekay
Yes 30th of this month, and yes that's when the segmented netdb stuff is absolutely due
dr|z3d
are we likely to meet that deadline? obscuratus?
obscuratus
dr|z3d: From my perspective, we're getting close. I have a handful of fixes to submit on top of what I pushed last week.
obscuratus
It seems like I always have some lingering issues, but as time has progressed, the list has become shorter.
dr|z3d
close is good, obscuratus. is it 30th of this month close? :)
dr|z3d
and how's the presentation looking?
eyedeekay
If it can't be then then we push back the release, it's not a big deal to do so really
dr|z3d
Regarding the console presentation, I figured we probably just want a single page for all the client dbs.
eyedeekay
Still quite imperfect but not entirely fugly
obscuratus
It also depends on what you call "major changes".
dr|z3d
well that's something. not entirely fugly is good. super sexy is better :)
obscuratus
I have one issue I'm looking at. There seems to be a non-obvious regression it how LS are published. I feel like I'll either have a fix or an assessment of this issue done by Wednesday.
obscuratus
I'm getting a lot of DSRM on LS lookup and LS store on my testing network. I really shouldn't be getting DSRM in such a well-controlled environment.
obscuratus
Or, maybe only occasionally.
eyedeekay
Does the increase have a predictable relationship with the number of subdbs in use?
obscuratus
I have a reliable reproducer using a single SAM server <-> client app. It looks like it is happening right away with only that single app (as well as the shared clients of course).
obscuratus
It's odd. If I put the server on a segmented netDb router, and the client on existing 2.3.0, everything works OK. But the client fails when on a segmented netDb router.
obscuratus
I think part of it is related to the issue that libsam3 is a limited SAM library. You can't even set a timeout. But, in this case, it is helpfully highlighting an issue.
eyedeekay
Interesting, thanks for examining it, sounds like it would be difficult to gain insight into this without the testnet
obscuratus
Yeah, I can play around with a lot of stuff on the testnet that would be really difficult on the actual network.
eyedeekay
The client is unpublished right?
dr|z3d
If obscuratus's tests lead to some improvements with the SAM lib, that's not altogether a bad thing.
obscuratus
eyedeekay: No, the client and server are published. (I thought to look at that also).
obscuratus
dr|z3d: Perhaps. I'm actually thinking of moving more to the i2psam c++ lib. I haven't played with it much, but polistern uses it for pboted.
eyedeekay
Which one is libtorrent using now?
obscuratus
They've pretty much rolled-their-own SAM interface, as near as I can tell.
eyedeekay
Yeah I was just skimming and coming to the same conclusion
obscuratus
Back to the segmented netDb issue I'm seeing, I notice in the logs that the 2.3.0 router really isn't even seeing any DSRMs, but the segmented netDb router is processing some DSRMs.
obscuratus
For the un-initiated, a DSRM is a DatabaseSearchReplyMessage. It's sent when a FF couldn't find the search that was requested by a router.
obscuratus
Oddly enough, things still seem to work, usually, even though the segmented netDb routers are wrastling with DSRMs. It would be easier if it was a full fail.
obscuratus
I2PSnark and ZZZOT all seem to be taling to each other well-enough.
obscuratus
It reminds me of a timing issue we had a few years back where your first search would fail, but subsequent searches would succeed.
dr|z3d
it wasn't a few years back, we had that last year :)
dr|z3d
if you need another SAM client to test with, obscuratus, i2pchat uses Java's native SAM lib.
obscuratus
dr|z3d: Good info. We'll need to test the segmented netDb stuff with as many clients as we can.
dr|z3d
*thumbs up*
dr|z3d
if you get it installed and need someone to chat to, pm me and I'll send over my i2pchat dest.
obscuratus
eyedeekay: I've pushed four more patches up to branch i2p.i2p.2.4.0-test1-obscuratus-wip-20230822 on my i2p.i2p repository on git.idk.i2p
eyedeekay
Thanks obscuratus I'm looking at it now