mygoditsfullofstars

So it turns out that a subsequent commit to the source after disabling GPU acceleration has unintentionally squelched the log spam (which from my testing seems to be caused if the bundled CEF GL-ES library is loaded), but there is still an underlying issue because CEF is falling back to using swiftshader which is a software renderer (CPU based). From the CEF source: It is controllable via the --use-gl flag - but its unclear why CEF has fallen back to swiftshader or why it is trying to use the EGL back-end to begin with - so what I said earlier still stands, this could still be a bug within CEF on linux itself, professor google also suggests that its not the first time CEF has had such issues on Linux. Anyway, if I discover anything else I'll post.

I've made an interesting discovery tonight about the log spam in question when GPU acceleration was enabled for CEF (and no, its nothing to do with my system or gpu type or drivers), I am discussing my findings internally with our devs to see what they think. If what I think is the issue turns out to be true I'll post back. The log spam errors are EGL related.. Why CEF was trying to use EGL on a Linux system is what I am curious about...

<redacted>

I believe he was trying to insult my intelligence by saying I am running it in a VM and the issue must be on my system (go and look at what he wrote and the tone). And he did the same thing when I mentioned that another team member had the same issue (it must be their system too and not an issue with CEF). I treat people how they treat me. Why should I change my attitude when he was the one being condescending and insulting to begin with? As for my role within the team, I was pointing this out to him so he can learn that he's not dealing with a noob who knows nothing (which is how he was treating me).

@Drakeo I am running the viewer on bare metal with the proprietary NVIDIA drivers. Stop making stupid assumptions. Go and look up who I am on the FS website (I'm not some clown who is just experimenting in a VM). I know better than to try and run it in a VM FFS. There is something inherently broken in this release of CEF and GPU acceleration on Linux. I am not the only one who saw the same GPU log error spam. Are you going to accuse the others of trying to run it on a VM too? Oh and before you even consider saying my hardware is old. Don't go there. Intel i9, RTX-3070 GPU.

No not really, personally I have a lot of items that were bought from vendors who no longer have their store and plenty of non copy stuff - also kindly point me to the redelivery function on the sl marketplace where everyone can recover their copyable items bought there. 2FA is opt in and unlikely to become mandatory. Entirely up to the user. Not sure why you have such a chip on your shoulder towards 2FA but whatever, your choice. I don't agree with the rest of your reasoning - done arguing with you.

You are also overlooking what I mentioned about an accounts inventory being purged if its inactive for long enough. Imagine someone deactivated a compromised account and its owner was on vacation or just having a break from sl for a bit. In some cases LL might not be able to restore an accounts inventory.

Really? Every account that I have terminated and then later chose to change my mind about has involved getting LL support involved to recover it. Unless they have made the process easier in recent years. Are you suggesting its a good thing that the delete page hasnt been protected?

It deactivates the account at first and involves having to engage LL support to try and get it back.. That page should be blocked by 2FA regardless. Say someone malicious manages to steal the current password to an account, then out of spite because they cant change the password or do other things to it, they delete the account. Then the legit owner of the account has to then go through all the hassle with support to try and get their account back (even worse, if its left deactivated long enough, the accounts inventory will be purged)... Easier and better to just protect that page to save someone having to go through the hassle. I was surprised on how LL implemented this actually, I would have thought they would just prevent account login entirely without 2FA, rather than just protect a subset of functions after logging in.

A Jira has been filed for this already on https://jira.secondlife.com/browse/BUG-231234 Hoping it gets addressed promptly.

Pity it doesnt protect the "delete account" page yet..

Yep, another team member ended up with a 99GB CEF log and it filled their partition... So its off for now, we can probably revisit turning GPU acceleration back on next time there's a CEF update for Linux

I am disappointed in what Soma have done too, at the same time I kind of get why (due to what I said earlier). And I was affected by this move on their part too, I had SomaFM streams on my parcel.. There's no reason I cant just open the stream in other software though (eg: winamp, vlc etc) and it will still work. Maybe LL could also look at disabling auto-play of music streams by default in their viewer too since this could in theory contribute to situations like this where someone has placed a stream on a busy region. Just about all the listener supported streams I listen to (not just Soma) have started to crack down on stuff like this since most people don't want to donate. Something has to give eventually. Looking at the SomaFM site, they need to clear around $40k USD each month to keep operating. I am betting only about 10% (at most) of the listeners actually donate.

Its disabled because its broken in the current CEF release for some reason. @DrakeoFeel free to compile and debug CEF and let me know what the issue is. This was making my logs about 2GB in size in no time.

Can I just suggest openly that people dont try and look for ways to circumvent the block in sl, it will only force Soma FM to take more drastic measures.

If MOAP gets abused to circumvent the block, I can almost guarantee you they will be blocking the CEF UA next (if they haven't already done so).

They are blocking FMOD specifically. A very simple test with curl proves that fact. MOAP they can do the same thing, block the CEF user agent string (which I believe has 'second life' in it somewhere).

From the GM of Soma FM in a response to my email to them: So, they are not punishing just sl users specifically. I was right, its a cost thing. And I am one of the guilty people who have listened to Soma FM for over a decade and never donated once. Something had to give eventually. People like myself are probably to blame.

I think Soma FM are trying to reduce operational costs.. It certainly doesn't help that the LL viewer autoplays streams by default out of the box... Say someone has a Soma FM stream set on a busy region, they would be seeing a lot of listener traffic.

Is it ONLY that region you have trouble teleporting to? No issue with teleporting to other regions whatsoever?

Its not just people with really low end systems that are affected by the regression in performance. I see this on my i7 with a GTX-1080 even on medium settings -- For example, On the non-eep viewer I went somewhere that had a fairly static scene that wasnt changing (also no other avatars around to render except my own) and then set the camera angle to default, noted the FPS (which was around 50), then switched to the EEP viewer that was on identical settings, even the same converted windlight, identical camera angle, and observed a 20FPS drop. All I did was a simple log out and log in between viewers and its reproducible every time. In general, eep just had this clunky and slow feel to it and I went back to non eep out of frustration.. There is something woefully inefficient going on under the hood if eep is managing to slow things down that much... Anyway end of my rant, my 2 pennys worth, EEP is not fit for release.

You should consider playing in the lottery

I am skeptical of this. I have experienced many ""acquisitions" in my working career and not ONE has had a good outcome in the end. The results ranging from restructuring and people being made redundant to making the place so sh** to work that people quit In droves.. And of course, the new owners always give the ""don't worry folks, it will be business as usual" speel to try and calm everyone down, but we all know that's usually a load of BS. The last "acquisition" i got tangled up in a couple of years back, we got the usual "stay calm" speech from the new owners and were promised the usual BS (nothing will change, business as usual etc), then not long after, suddenly almost everyone in the company had to reapply for their jobs. I had been there for over 6 years at that stage. everyone was like wtf? So I was interviewed via video to some crowd of suits at a table on the other side of the country, almost everyone was. Not long after that, the "new owners" kindly told everyone that they were out of a job in about a months time and they would be running the company out of a small office on the other side of the country. Meanwhile, they flew in people from the other side of the country for me to train in doing my job. It was pretty soul destroying. Only 1 person survived the cull, the head infrastructure engineer, and it was just to use him so that they could wind down all the data center stuff. That's just one "acquisition" experience I have had, the rest I have experienced in my working career have had equally disappointing outcomes, the new owners come along and ***** everything up royally (the tech itself and/or the working environment). You cant blame your userbase for having doubts about a good outcome here.

There HAS been an increase in this over the last couple of months for me. Its at the point where its seriously annoying again. I think the original issue (someone can correct me if I am wrong) which was supposedly fixed, was some race condition in the simulator software? When the original issue was declared fixed, things went back to being good again for me, really good, failed tp+disconnect was rare. Now its happening often again! I suspect there is an issue again and latency is a probably an influencing factor. In some cases users will naturally have a higher latency due to geographical distance to the network that hosts the simulators. For me, the latency to the US is around 180ms-250ms (depending on network conditions and where in the US the host is). The thing is, LL will look at this and say "we see no issue" - my response "now add some latency to simulate your overseas visitors who don't have your ultra low latency times like you do in the US and see how you go!" -- and that's easily achieved from the US, use an overseas VPN to conduct tests. It is absolutely not my internet, as I recently changed ISP's and broadband solution types (from HFC cable to Fiber to the home). The issue was no better or worse after the change. It is unlikely to be my router either (its no potato I assure you, its a Cisco (current and supported model) with a list price of around $10k in the US). Everything is ethernet too, no patchy wifi garbage.