X-Plane 11.20 is, of course, final! Like all good Klingon software, it was not so much released – rather it escaped captivity, leaving a trail of blood in the path of all whom opposed it.
I am aware of two plugin-related issues we are tracking:
Some legacy plugins with widget UIs are missing buttons when the UI is at 150% or 200% scaling. We have a fix for this, but for now you can work around the problem by turning off UI scaling in the graphics settings tab.
Some Mac plugins compiled against libstdc++ crash with the Steam version (but not the non-Steam version).
Philipp and I are still discussing what to do about this second thing, but if your plugin is in this category (so far we’ve seen it with HeadShake and X-Ivap), my suggestion is: compile and link against clang’s libc++ on OS X – it’s the native Mac C++ runtime and the one that’s going to work well in the long term.
We’ll release an 11.21 release candidate either late this week or early next week, once we collect the bug fixes that got away.
X-Plane 11.20r2 has only one bug that we are trying to fix: a bug in the plugin SDK that can cause some plugins to crash when creating and destroying windows.
Tyler and I dug into this and found that the fix was a bit more intrusive than we wanted for this late in an RC.
So: if you are a plugin developer and you are working with the new SDK, either for VR compatibility, to use the new 3.x APIs, or just because you are updating the plugins, and you have time this weekend, please email me and I can send you our new fixed XPLM DLLs.
My hope is to get half a dozen plugin developers to bang on them over the weekend, so that when we cut 11.20r3, it really will be the release.
We received a last minute bug report that X-Plane 11.20r2 deletes scenery packs from scenery_pack.ini. This is true! It is also by design, not a bug, the same as X-Plane 10, and not particularly brilliant on my part.
scenery_packs.ini persists the order of your packs (putting newly discovered packs “in front” in alphabetical order upon discovery) and maintains enable-disable status. But it does not retain any other information. The file is totally rewritten on every run and the following otherwise useful stuff tends to get destroyed in the process.
Comments and notes to yourself
Whitespace and formatting
Any scenery packs that can’t be found
Unfortunately this means that if you symlink to an external drive that is unmounted and run X-Plane, your scenery pack order gets lost. This definitely does suck.
In a future version, I can fix this. But we’re not going to mess with it now during RC because it’s totally not changed or broken compared to all past versions of X-Plane since we introduced the scenery_packs.ini file.
For now: you can lock the file as a hack-around to preserve your well-created order while your remote drive is unmounted – when you re-mount and restart, your order isn’t lost. But in this order, new packs aren’t persisted to the file, although they will be used.
If there’s things you want the scenery_packs.ini file to do, commenting on this post would be on-topic. One thing to consider: if we can’t drop missing packs, how do we know a pack was uninstalled? How does the file ever get cleaned?
In the X-Plane 11.20b1, we’re shipping a web browser for the first time. We’re using the Chromium Embedded Framework (CEF)—essentially the same guts as Chrome, wrapped inside X-Plane.
For the time being, it’s being used in one place only*: to support in-app upgrades, so that if you have the demo, you can buy the full version of X-Plane without having to go to the web site. The web view is seamless—you can’t tell by looking at the app that it’s not just part of the native X-Plane user interface.
While its present use is quite limited, if all goes well, we’d like to expand its use elsewhere in the sim—in The Glorious Future™, we could potentially use it to load online charts and the like.
But, there’s a hitch!
There’s not a good way to load two copies of CEF at a time. And, since some plugins (like the ones used in the new Flight Factor A320 Ultimate) depend on a version of CEF provided by a (global) plugin, we have to disable X-Plane’s web browser functionality in the presence of that plugin. This isn’t a big deal right now—after all, if you’ve installed payware, you probably already own the sim anyway—but it will be a shame if, in The Glorious Future™, you have to choose between your payware and core X-Plane functionality.
The situation is even worse than you might expect: CEF absolutely doesn’t support multiple initialization calls in the lifetime of an app—you can’t initialize it, shut it down, then re-initialize it. That means it’s not even possible for plugin authors to be good citizens and “relinquish” CEF to X-Plane, such that you could at least use X-Plane’s browser whenever you’re not using an add-on that depends on it. It’s all or nothing! In fact, you’ll have to uninstall the CEF plugin before X-Plane can use a web view itself.
There are two possible fixes here:
At some point in the future, we provide access to CEF via the plugin SDK. If we did, it would have to be to the C API only—no fancy C++ wrappers for you. Speaking as someone with the memory of writing for the C API fresh in his mind, let me say: this isn’t a whole lot of fun, and it’s rife with the potential for memory leaks. Moreover, this would break any existing addons that depend on CEF—they would be forced to migrate to the SDK version of the API.
We could (theoretically) compile an X-Plane-specific version of CEF that would not conflict with the version plugins are using. This would require renaming all the Objective-C symbols on Mac, and renaming the DLL on Windows. I’ve not investigated this for feasibility, but it’s certainly theoretically possible. This would allow X-Plane’s version of CEF to coexist with (one) plugin-provided copy, albeit at double the CPU and memory cost.
Having us expose CEF via the SDK is not an unequivocal win. CEF is notoriously incompatible between versions—you can more or less guarantee that even minor updates will break compatibility of the API somewhere. That means X-Plane would be stuck at a fixed version of CEF for at least the lifetime of a major version to prevent breaking plugins. X-Plane 11.20 uses CEF release 3202; it would be at least the next major version before we updated CEF to a newer version. There are two major downsides to this:
When X-Plane updates to that newer version, it would break plugins compiled against old versions of the SDK that depend on CEF. That’s a lot more aggressive than our normal deprecation policy, and it makes CEF plugins a potential maintenance headache for plugin devs.
If your add-on really, really needs features from the latest and greatest CEF release, you’re out of luck—wait a couple years (!!) and maybe we’ll update.
So, here’s what I’d like to hear from plugin devs:
Are you currently, or are you planning to use CEF in the future?
Would you be willing to use the C version of the CEF API only?
Would you be able to accept the limitations and potential headaches (outlined above) of X-Plane-determined CEF versions and compatibility?
If you don’t mind telling me a bit about your intended use cases, that’d be very helpful as well!
EDITED TO ADD: There’s a third option here that I didn’t consider: X-Plane could provide a “wrapper” around CEF that offers “just” a browser surface, and simple interfaces like the ability to change the URL programmatically. This has the advantage that we could update CEF regularly without breaking the API (though there’s always the risk that a new version of Chromium would change how it handles your HTML + CSS + JS). The disadvantages are twofold:
a) If a plugin developer really, really needs the full power of the CEF API, they’re out of luck (or we’re back to square one with respect to having to disable core X-Plane functionality in order to support the plugin).
b) New XPLM APIs are a massive tax on our development time. Every time we need to make a change, we have to go test a dozen plugins… then go through an extensive beta… then inevitably hear about a show-stopping bug we introduced three hours after a release goes final. 😀 In contrast, providing a (major-version-stable), transparent copy of CEF costs us very little dev time; time that would otherwise be spent maintaining the XPLM API can be spent on, like, major features.
* Aside: This is actually how we test a lot of new tech in X-Plane: we find a single, relatively out-of-the-way place to make use of it, ship it in a major version update, and wait to see if it blows up. Betas catch a lot of bugs, but not all, so this is a way of hedging our bets when it comes to code that hasn’t been battle-hardened. This is how we worked the bugs out of the user interface framework (Plane Maker’s panel editor was the testbed), the X-Plane 11 particle system (behind the scenes, it was used for some very minor effects in X-Plane 10.51), and more.
X-Plane 11 has had several fantastic updates to the library of lego-brick airport objects that ship with X-Plane. We’ve had new European-style objects, elements for small airports, and the terminal and jetway kits let you build really great looking airports that fit the exact shape of the real-world airport you are modeling.
In that context, I’m pleased to show you a sneak preview of the next set of lego-brick objects for gateway airport authors.
You can download a preview if the new lego bricks here.
Fellow nerds^H^H^H^Htrekkies^H^H^H^H^Hnerds are aware of the even-odd phenomenon – for nine films, every other Star Trek movie was pretty bad, then the next one would be okay. VR previews are now like that, with VR2 and VR4 being fatally broken. We’re planning to just skip 6 entirely.
Anyway, I broke VR4 – my code to put a loading screen in the Oculus Rift had a side effect of killing the render resolution on the HTC Vive and WMR-based headsets. Since I only have the Rift (Chris has one of everything) I never saw the bug; since I thought my change wouldn’t affect anything outside the Rift, Chris didn’t re-try the Vive.
So, VR5 is now up – this should restore the Vive/WMR headsets to the performance of VR3.
We can confirm that ASW is not working in VR4 or VR5 – we’re looking at this now, but for VR4 and VR5 you’ll see ATW but not ASW, regardless of Oculus debug settings.
(Edit: the first version of this post confirmed no ASW in VR5. ASW is off in both VR4 and VR5 because the native rift SDK is not permitting ASW for X-Plane. A number of commenters were concerned that they would lose ASW by going from VR4 to VR5. You will not! ASW is already off in VR4; if you like VR4’s performance, you clearly don’t miss ASW.)
(Edit 2: We have a very promising lead on the ASW issue for Oculus. We have it working again here but I want to do more testing before I declare victory. We do not need any more information. Nothing you can do on your end will fix this. If you think you’re fixing it by “jiggling the handle” you’re probably just falling into placebo effect. If everything goes well we will have an update to fix this over the next day or so).
****EDIT: There’s apparently an issue for people running the Vive and WMR where they’re seeing reduced resolution. We’re looking into it and will post an update as soon as possible.
****EDIT2: We’ve found the issue affecting Vive and WMR. We’re testing a fix internally and will release an update hopefully in the next 24 hours. Please do not submit any more bug reports about Vive/WMR resolution.
11.20 VR4 is Live on the servers. Aside from the usability fixes that Ben already mentioned, the major ‘feature’ in VR4 is…Oculus users will no longer need SteamVR. If you downloaded it just for X-Plane, go ahead and remove it. It will no longer be necessary.
As we said we would do from the very beginning, we investigated the relative performance of X-Plane through the native Oculus SDK versus SteamVR and what we found, through data collection,is that the overall experience for Oculus users was better by going through the Oculus SDK directly. I know many of you are thinking “Duh! I told you that a month ago ya big dummy!” and yes…yes you did. Fortunately/Unfortunately, we try not to make engineering decisions based on gut feelings and anecdotal evidence when we have a way to collect actual numbers. We wanted to tackle a majority of the usability issues affecting everyone before we looked into performance.
In the various A/B tests that we performed, we found that going to the Oculus SDK directly got us about 25% improvement in frame rate. This does not necessarily indicate that there’s anything wrong with SteamVR itself. There are several factors influencing the performance in VR. First, Oculus has their “home”, that little bachelor pad where you hang out while waiting for games to load. SteamVR has their “home” as well. When you use SteamVR, BOTH are running on your machine. Those houses are not free and X-Plane is already CPU bottlenecked so anything consuming CPU resources is going to directly affect performance. (I noticed an Autodesk updater in my task manager that was stealing 5% of my CPU consistently. That too was decreasing my performance….every bit matters!). Going directly to the Oculus SDK removes the SteamVR house from the equation.
Sure, getting 25% improvement is a huge win, but that’s NOT the biggest win. The biggest win, in my opinion, is that Asynchronous Space Warp (ASW) works MUCH better even at very low frames rates down to about 22.5fps. It appears as though the timing of the frames is critical for ASW to work properly. Being at 22.5, 30, 45, 90fps feels smooth! Being in between those frame timings seems to make ASW lose its mind creating an annoying judder; the opposite of what ASW is supposed to be doing for us. Oculus seems to be V-Syncing us to hit those intervals, allowing their algorithms to make reliable timing decisions and predictions. It’s my suspicion that SteamVR was just not hitting those intervals, causing ASW to flip out.
TLDR; Performance for Oculus will be on par with what Vive users have been seeing all along. The smoothness of the rendering seems consistent even down to 22.5fps. If you’re a Vive user, you will still, of course, need SteamVR as that IS your native SDK. If you’re a WMR user, you will still need SteamVR. I have not seen any reprojection issues with WMR like we have with Oculus. Supposedly the upcoming versions of SteamVR have some performance improvements coming for WMR users as well so we’ll be sticking with SteamVR for all headsets other than Oculus. That can always change in the future…based on data.
We’re starting internal and private testing of VR preview 4 – if it goes well, we’ll release it early next week. A few notes on usability:
Have Your Mouse Cake and Eat It
We got a metric ton of “bug reports” that users couldn’t click “Disable VR” when using the 3-d mouse. I put bug reports in air quotes because disabling click zones on the main monitor when using the 3-d mouse was totally intentional! (In other words, I broke that button on purpose.) My thinking was that you might click on “Disable VR” by accident while in the headset because you don’t know what the 2-d mouse is hovering over while clicking.
In VR4, we have a solution to the problem of whether the 2-d or 3-d mouse is the “intended” cursor when a click happens: we read the headset’s “on the user’s head” sensor and disable the 3-d mouse (temporarily) when you take the headset off. So you can be clicking in 3-d, take off the headset and immediately click “Disable VR” and we figure it out. This feature is great for developers who need quick access to 2-d debug tools like the texture browser or DataRef editor. Read More
Plugin developers: the title says it all – don’t register a drawing callback that doesn’t actually draw anything.
When I write it like that, it sounds a little bit silly, but some plugins register a drawing callback on startup, because they might draw something – but then only do the drawing when certain things happen. I am guilty of this myself – libxplanemp registers its drawing callbacks at init time even if there are no multiplayer planes to draw.
Most plugins did this for simplicity and to avoid any strange rules about when you could register things – the original 1.0 SDK was a lot pickier than what we have now. And back in 1537 when the French invaded Flanders, the cost of a no-op a draw callback was pretty cheap – just the cost of the function call-out and a few instructions of book keeping.
Over time, as X-Plane’s renderer has become more complex, this has become not the case. X-Plane now does some significant work to save and restore GL state when we hit a drawing callback, and that overhead is going to become significantly more expensive when we move to Vulkan/Metal.
To optimize this, Tyler modified the drawing callback code in 11.10 – starting with 11.10 if we find that no plugin has registered a given drawing callback, we skip the entire process, including the OpenGL book keeping.
So if your plugin goes in and registers a callback for every single drawing callback type then we have to fall back to the slow path, carefully stashing our GL state and prepping for your running at each drawing callback point.
Here’s my guidance:
Never register for a drawing callback you’re not going to use.
Do delay registering for a drawing callback until it is possible to need it.
Don’t try to optimize your drawing callbacks by registering and de-registering per frame.
So as an example, it would be reasonable for a multiplayer API to register a drawing callback when the first airplane enters the system and then keep it around until the last airplane goes away, even if that airplane is off screen.
(The reason for that last rule is: if we have to make an expensive graphics transition once we discover that drawing callbacks will be needed, we don’t want to get thrashed by drawing callbacks disappearing and then re-appearing over and over.)