Serious Development Barriers

10 May 2020 17:18 - 10 May 2020 17:19 #46481 by Algernon
So, during the FlightNight last night, I was disturbed to hear from Alex and Geed that their EF2000s had some serious problems with their flight characteristics, which was a complete surprise to myself and Porcius, who have never experienced them at all. As an experiment, Mav flew it and found the same problems. It would appear the aircraft is behaving differently on Windows and non-Windows systems.

My first sense was one of humiliation - I've put hundreds and hundreds of hours into creating what I think is a pretty decent aircraft, for an amateur, only to find that when my friends fly it on a different OS, it's embarrassingly substandard. I had no idea this was a possibility, and frankly, I was super pissed off.

I wouldn't know where to start trying to pick it apart in order to submit a ticket. And even if I did, I doubt I could make anybody care - it appears that there are some seriously declining standards in FG releases now, as 2020.1.1 RC has been put out with a known issue which, with a single click, can ruin your entire flight. It's not an oversight - I'm told the problem has been known about for quite a while - and I think that's pretty major, and pretty damning, almost beyond belief.

Both this, and that stupid MP model reloading thing, appear to be an attempt to manage system resources, done very badly, at the expense of the user experience. Most of us know if you run a Flight Sim, you need a system that can handle it - I bought my machine especially for FlightGear - and it has ample resources. So why do I have to put up with MP aircraft turning into gliders the moment I stop looking at them? And now, suddenly losing all scenery if I use the new, weird MP view, which is far from finished as a concept and replacement for model view. It's a poor show, and getting steadily worse.

So, dear developers, please be aware - unless you're able to run three different OSs, and have the immense time it would take to tune and FDM/FCS across all three of them, you might find that pilots on a different OS to yours might not think so highly of your work as you do.

Please Log in or Create an account to join the conversation.

10 May 2020 21:02 #46490 by enrogue
I feel your pain Algy :)

I'm quite nervous about the general response to the F-5E as well even though I haven't done nearly as much work on it as you have done on the EF2000

I'll respond to a couple of points:

- The issue with the MP gliders appearing & disappearing I only noticed on the flight night because I usually have my MP Aircraft LOD set to 'High Detail Only' because MP carriers used to require this (I don't know if they still do). My settings were recently completely reset/destroyed due to an issue I was trying to track down, so I noticed it & immediately (during the flight in fact) reset my MP aircraft setting to high detail & fixed it

The idea of the Low/High detail is a good one - when aircraft that are hard to see in the distance it makes sense to load a low detail model, but if the settings are wrong or if there is *no low detail model* like the F-5 (it's CC so there won't be unless someone makes one from scratch) you get the glider

I now have a fairly decent computer as my FG machine, so I now am fine with high detail only models

The issue with the terrain unloading has been around for a while & has been complained about on the forums for quite a while, but really only brought to the attention of the mailing list recently - it was initially thought of as a LOD paging issue with OSG, but StuartB found an issue with the terrain tiling and has literally *just* pushed a fix

I just checked it out & compiled & it did make a noticeable difference for me, but I've only tested it on my fastest machine (1TB NVME SSD + 32GB RAM means terrain gets cached by the OS & brought back fairly quickly - probably why I didn't think it was that bad in the 1st place!)

I'm going to compile & check it on my much older & slower iMac - lets see if it's better (the iMac gets its terrain from a network drive, so the difference should be noticeable!)
The following user(s) said Thank You: Porcius

Please Log in or Create an account to join the conversation.

10 May 2020 21:07 #46492 by StuartC
F5 works as intended from what ive seen. Some small fine tuning of bits but nothing major is needed for it.
The following user(s) said Thank You: enrogue

Please Log in or Create an account to join the conversation.

10 May 2020 21:53 #46494 by enrogue
Just compiled & tested on the 2013 iMac - not the slowest machine in the world, but 7 years old

Some other recent bug-fixes have actually made this machine usable again

This change just pushed to the terrain tiling makes a noticeable difference - I spawned at EGOY & just within range 2 players were at EGOV - once I switched to them it loaded the terrain & I could switch views & switch between them without it reloading the tiles - a major improvement

So a big win there.

The model detail viewing I would recommend setting to High Detail only if you have a decent machine - especially for camera work. I know Richard has been doing some nice low poly models with single texture atlases derived from fgaddon for inclusion in FGdata, but that isn't going to happen for non GPLed models :)

I know there was another issue you mentioned with the view, but I can't remember what it was - if you could remind me I could try & track it down & log a ticket
The following user(s) said Thank You: Algernon

Please Log in or Create an account to join the conversation.

11 May 2020 07:38 #46498 by Algernon
Thanks! Some good suggestions there.

I probably laid it on a bit thick because the issue with the EF2000 really hit me pretty hard. In a more reasonable state of mind, I do understand how the scenery tile issue thing has slipped through into the RC, and somehow it had never occurred to me that you could only see a low quality model if there was a low quality model available!!

More to follow...

Please Log in or Create an account to join the conversation.

11 May 2020 12:41 - 11 May 2020 12:43 #46504 by Richard

Algernon wrote: it appears that there are some seriously declining standards in FG releases now, as 2020.1.1 RC ... and pretty damning, almost beyond belief.

I'm responding to what you originally wrote; I realise that probably you were a bit grumpy at the time.

It is highly recommended that anyone developing aircraft for FG should at least read the dev list to keep up with what's going on. The official forums don't have the same level of detail - participating in the devlist is also the best way to get things changed.

Since it seems that the new version system is still causing confusion I'll explain again, sorry if you already knew this.

In 2018 James proposed changing the way that releases are made from the frequent (upto four per year) to instead focus on producing a stable version which would then be maintained until the next stable version. This happened with 2018.3 when there was a feature freeze and solid concentration on finding and fixing bugs before releasing of 2018.3 LTS (long term stable). Since release there have been updates to the 2018.3 LTS to fix bugs that have since been found - but no new features. The other feature of the LTS version is that the (FGAddon) aircraft are also maintained so that only bugfixes are applied. Obviously this doesn't apply to FGUK releases because these come from this site.

You can expect some problems with 2020.1 - however one of the main points of this release is to work towards a feature freeze for 2020.2 (which should be the next LTS).

Once 2020.2 LTS is released it should be stable and of a lot higher quality than the current 2020.x veresions.

It is my recommendation that all aircraft developers should start at a minimum start developing with 2020.1 (or whatever version is the feature freeze version) - to find and fix any problems with their craft but also to take advantage of new features.

Algernon wrote: I wouldn't know where to start trying to pick it apart in order to submit a ticket. And even if I did, I doubt I could make anybody care

You're definitely not going to make anybody care if you don't tell them about the problems; e.g. raise a ticket or post to the dev-list. It's easy to raise a ticket and the ticket could simply be a list of things that don't work properly and someone will investigate.

I've just tried the EF2000 on 2020.x, 2019.2 (nightly), 2018.3.5, 2017.2 and I can see a few issues between 2020.x (compositor) and 2018.3.5 and I found the same handling problems in all versions but then trying to do unstable delta wings using yasim is not easy.

Algernon wrote: Both this, and that stupid MP model reloading thing, appear to be an attempt to manage system resources, done very badly...It's a poor show, and getting steadily worse.

You're misunderstanding the way that the high detail and low detail models are supposed to work and the relationship between fallback models and the LOD system.

This is something that I've explained before both in the forums here but also verbally on flight nights. for the 2018.1 release I had this to say (see fguk.me/forum/flightgear-general-chat/76...8-3-1-released#41141 )

Richard wrote: Fallback models are actually really neat; and something that FGUK should support as it means you get to see something - the fallback models will load very quickly; and then the high detail will switch in.


The LOD system is explained on the devlist - here sourceforge.net/p/flightgear/mailman/fli...ail.com/#msg36335571

Fallback models are documented here wiki.flightgear.org/MP_Fallback_models

All models should set a fallback model index. If there isn't a suitable model in the list pick the closest match, submit a new model.

Because the fallback models ship with FlightGear all you need to do is to set the property /sim/model/fallback-model-index (from your -set.xml) to something suitable from the file (eurofighter is 503, F-5 is 525); and then you don't even have to have the full aircraft model installed to see something.

Remember that all models will have two models loaded, the low detail and the high detail. Models that don't set a fallback model index will default to the glider.

Find the fallback model index by looking in $FGData/AI/Aircraft/fallback_models.xml ( sourceforge.net/p/flightgear/fgdata/ci/n.../fallback_models.xml )

If you don't like this feature then simply set "High Detail Only" from the Static LOD dialog and it will work as before

Algernon wrote: unless you're able to run three different OSs, and have the immense time it would take to tune and FDM/FCS across all three of them


It is highly unlikely that any FDM/FCS problems are platform specific, except possibly those related to case sensitive filenames. What is much more likely is FDM/FCS that shows problems on different systems due to frame rate or different pilots.

Please Log in or Create an account to join the conversation.

11 May 2020 14:17 #46505 by Algernon
Thanks for taking the time, Richard, and patiently dealing with my... shall we say, ill-informed vent.

With the LOD model thing, yes, I totally misunderstood how that works and therefore should probably have made sure I knew what I was talking about before I slagged it off.

It is highly unlikely that any FDM/FCS problems are platform specific, except possibly those related to case sensitive filenames. What is much more likely is FDM/FCS that shows problems on different systems due to frame rate or different pilots.

That's what I would assume too, and have been thinking about today - and it's possible that there is a frame rate issue because some aspects of the FCS are switched using Nasal (though I've tried to keep that to a minimum). I know how tricky it is to do a delta wing, and especially one like the Eurofighter, in YASim, but I really thought I had cracked it (with the help of a physicist we used to have as a member) years ago.

Anyway, I am calmer now and regret my outburst - particularly as I am an unorthodox developer and know full well I don't help myself by not staying up to date.

With regards to raising tickets, I am trying to do this, however I do like to try and gather as much information as I can before I do it. It's also requested that before raising a ticket, you have a discussion in the main forum first, to see if it's an isolated issue or has been noticed by others, which I'm also trying to do now. I will try and get better at this.

Basically, I feel like the EF2000 needs a reboot and so do I, as an aircraft developer. Part of the new project - Typhoon Tranche 3 - will be to try and fall better into line with the FG direction of travel in terms of the software and its latest features (Compositor, Canvas); I will also try to improve my workflow as you suggest.

I'll also take note of the versioning thing, though I must say, it does strike me as problematic to have release candidates which aren't recommended for stability, but are recommended for aircraft developers. I wonder if it might be perhaps more sensible to have two major branches, named to indicate whether it's the better version for pilots or for developers?

Anyway, again, thanks for taking the time and patience. There's nothing like realising where you're wrong for throwing cold water over a simmering rant! :)

Please Log in or Create an account to join the conversation.

11 May 2020 14:42 #46506 by enrogue
Richard: thanks for the reminder about the fallback model index - I had completely forgotten about it

just updated the F-5E in git, will update the zipfile in a bit

Please Log in or Create an account to join the conversation.

11 May 2020 21:31 - 11 May 2020 21:32 #46528 by Richard

Algernon wrote: ..it's possible that there is a frame rate issue because some aspects of the FCS are switched using Nasal
...
I know how tricky it is to do a delta wing, and especially one like the Eurofighter, in YASim, but I really thought I had cracked it

Mixing Nasal and or using property rules is not a good idea for and FCS because property rules run at the display frame rate

See the comments in sourceforge.net/p/flightgear/fgaddon/HEA...ms/m2000-5-yasim.xml - I maintain that it's not really possible to an unstable delta wing in yasim because of the way that yasim works.

There might be more future it using the same techniques that we did to model the Mirage-2000 using OpenVSP and JSBSim. I did the aero and Josh did the FCS and after almost 3 years I think we got to something that is close to being plausibly correct throughout the envelope. However I'm trying to not take on any more aero model projects until I've finished the ones that I've started.

Algernon wrote: ..raising tickets....It's also requested that before raising a ticket, you have a discussion in the main forum first,

That's good to hear. For example the z-offset post you made tonight is a candidate for posting to the dev list or raising of a ticket particularly as the view system has been under maintenance recently.

I don't think you need to discuss tickets on the forums; that's really for non developers; pretty much anyone in good standing here can almost certainly just raise tickets.

Algernon wrote: Basically, I feel like the EF2000 needs a reboot and so do I, as an aircraft developer.


Whilst I can understand why it feels like that I don't think you need to 'reboot' the EF2000. It's a good aircraft model and the biggest thing it needs is canvas displays. Canvas can seem very intimidating - but it isn't as hard as it seems. I can help to wire some basic displays and help with the learning process. Possibly the one thing to do is to try to organize a team effort

Algernon wrote: I'll also take note of the versioning thing, though I must say, it does strike me as problematic to have release candidates which aren't recommended for stability, but are recommended for aircraft developers.


Effectively that's what the two versions are; the LTS version is for pilots who just want to fly and the other one is for developers and those who enjoy new features at the possible cost of stability (i.e. like FG used to be).

Also worth saying that from now until the LTS release the focus of core development will be bugfixes and stability so now is a really good time to report bugs that are present in 2020.1 as hopefully we can get them fixed before 2020.2 LTS is released.

Please Log in or Create an account to join the conversation.

Time to create page: 0.170 seconds
Powered by Kunena Forum

Latest Forum Posts

PM Notifications

You are not logged in.

PM Mailbox

You are not logged in.

Latest updated downloads