[KSP 1.3.0->1.5.*] Mod Pods

[HebaruSan]

I suspect that some aspect of these tests is exposing a stock bug. For example, there are pretty rigorous reports of a large performance difference between launching from VAB and launching from the space center menu launch pad popup:

So we know that KSP has a bug in which the CPU can be tasked with doing extra, unnecessary work during launches depending on exactly what you click to get to the launch pad. Maybe some of that extra processing interferes with atmospheric calculations (e.g., applying drag forces twice)? If the running of the current test case looks like:

1. Reset to stock
2. Build craft in VAB
3. Click launch from VAB
4. Run test
5. Exit
6. Install ModPods
7. Click launch from space center menu launch pad popup using same craft
8. Run test

... then we're already in the domain of other factors potentially affecting the results even though it seems like they shouldn't. (I don't have the references on hand, but I think I've seen similar reports for fresh launches versus reverts.)

Obviously this is speculation, but it's consistent with what's been said so far, and the issue has to be something along these lines to account for a few unused parts seeming to affect unrelated parts of the game.

[TiktaalikDreaming]

2 minutes ago, HebaruSan said:

I suspect that some aspect of these tests is exposing a stock bug. For example, there are pretty rigorous reports of a large performance difference between launching from VAB and launching from the space center menu launch pad popup:

So we know that KSP has a bug in which the CPU can be tasked with doing extra, unnecessary work during launches depending on exactly what you click to get to the launch pad. Maybe some of that extra processing interferes with atmospheric calculations (e.g., applying drag forces twice)? If the running of the current test case looks like:

1. Reset to stock
2. Build craft in VAB
3. Click launch from VAB
4. Run test
5. Exit
6. Install ModPods
7. Click launch from space center menu launch pad popup using same craft
8. Run test

... then we're already in the domain of other factors potentially affecting the results even though it seems like they shouldn't. (I don't have the references on hand, but I think I've seen similar reports for fresh launches versus reverts.)

Obviously this is speculation, but it's consistent with what's been said so far, and the issue has to be something along these lines to account for a few unused parts seeming to affect unrelated parts of the game.

I'm leaning towards a similartype of explanation.  But it would be nice to tie down what's triggering it.

It would be worth while deleting one part at a time to see if the issue related to a specific part existing.   Because while it may be a stock bug, it would be nice not to trigger it. Or at least to understand when it's going to be an issue.  Hell, if the Devs are still diagnosing the bug, this might be more info.

So, @tex_nl, of you could try deleting parts, one at a time, probably starting with the service module mentioned as having dodgy antennas, that would be awesome.  If you have time and remaining care factor.

[Tex_NL]

19 hours ago, TiktaalikDreaming said:

I'm leaning towards a similartype of explanation.  But it would be nice to tie down what's triggering it.

It would be worth while deleting one part at a time to see if the issue related to a specific part existing.   Because while it may be a stock bug, it would be nice not to trigger it. Or at least to understand when it's going to be an issue.  Hell, if the Devs are still diagnosing the bug, this might be more info.

So, @tex_nl, of you could try deleting parts, one at a time, probably starting with the service module mentioned as having dodgy antennas, that would be awesome.  If you have time and remaining care factor.

I'm not sure when exactly but I'll look into it over the weekend.

[Tex_NL]

@TiktaalikDreaming
I am currently still in preliminary testing but as things look right now I can confirm it is something in the ../GameData/ModPods/Parts/Mk3to4 folder.
I started with a full install and consecutively removed sub-folders until I started noticing a difference. That difference came after removing Mk3to4.

After this round of testing RL has to take priority. I'll post my findings and detailed reproduction steps later.

Edited February 10, 2018 by Tex_NL

[TiktaalikDreaming]

26 minutes ago, Tex_NL said:

@TiktaalikDreaming
I am currently still in preliminary testing but as things look right now I can confirm it is something in the ../GameData/ModPods/Parts/Mk3to4 folder.
I started with a full install and consecutively removed sub-folders until I started noticing a difference. That difference came after removing Mk3to4.

After this round of testing RL has to take priority. I'll post my findings and detailed reproduction steps later.

Cewl.  Thanks.  

Even though I can't seem to reproduce this, I assume you won't be the only one.  It really helps to have someone willing to help chase down the issue.

[Tex_NL]

17 minutes ago, TiktaalikDreaming said:

Cewl.  Thanks.  

Even though I can't seem to reproduce this, I assume you won't be the only one.  It really helps to have someone willing to help chase down the issue.

I find it strange you are unable to reproduce it. But for now I am just glad you're taking this serious.

The dV at launch in my test rocket has not changed for all my tests (21 in total. 3 flights for 7 different tests) so it can't be anything related to weight, thrust or ISP. I get the idea it has something to do with drag and is too subtle to notice without automation.  During the flight MechJeb will throttle back to limit max Q. I still need to confirm it but I thought I noticed MechJeb throttling back more without Mk3to4 and therefore save fuel.

Will keep you informed.

Edited February 10, 2018 by Tex_NL

[TiktaalikDreaming]

4 minutes ago, Tex_NL said:

I find it strange you are unable to reproduce it. But for now I am just glad you're taking this serious.

The dV at launch in my test rocket has not changed for all my tests (21 in total. 3 flights for 7 different tests) so it can't be anything related to weight, thrust or ISP. I get the idea it has something to do with drag and is too subtle to notice without automation.  During the flight MechJeb will throttle back to limit max Q. I still need to confirm it but I thought I noticed MechJeb throttling back more without Mk3to4 and therefore save fuel.

Will keep you informed.

I haven't been testing with MechJeb.  I (this makes zero sense) wonder if it's something mechJeb related.

My test routine has been a stock rocket, aiming at some daftness in the aero to increase drag, and just SAS at go Up, and full throttle and see how far it gets.  I might try the same tests with (switching to liquid rockets) MechJeb

[Tex_NL]

22 minutes ago, TiktaalikDreaming said:

I haven't been testing with MechJeb.  I (this makes zero sense) wonder if it's something mechJeb related.

My test routine has been a stock rocket, aiming at some daftness in the aero to increase drag, and just SAS at go Up, and full throttle and see how far it gets.  I might try the same tests with (switching to liquid rockets) MechJeb

It is not entirely impossible MechJeb is to blame. But I do think it is highly unlikely as both 'clean' and 'contaminated' flights have been controlled by MechJeb.
I too have tried the simple 'SAS on and full throttle' approach with little to no difference in Ap.

[Tex_NL]

And now for the results of my first round of testing. I've tried to be as methodical, scientific, unbiased and reproducible as possible.

All flights have been launched from the VAB with the same 100% stock 2-stage rocket (see image below) and have all been controlled by MechJeb2 under the same conditions (also see image below).

All flights have been flown three times to average out any inconsistencies in MechJeb. (There weren't any inconsistencies. All dV results were within 0.5 m/s)
All flights started with the exact same fuel, weight, dV and TWR. The only mods involved in this test are MechJeb2-2.7.0.0, MechJeb Embedded Universal, ModuleManager.3.0.1 and ModularPods-1.0.9. The only chances between fights have been made to ModPods. The other mods all have stayed identical.

Flight #1 started with a full ModPods install. Flight #2 had only the TDProps folder removed. Each consecutively had one more sub-folder removed until none was left.

	Changes to ModPods	Average dV left/short over three flights.
1	Full ModPods	-309m/s
2	ModPods -TDProps	-309m/s
3	ModPods -TDProps -Decouplers	-309m/s
4	ModPods -TDProps -Decouplers -Mk2to3	-309m/s
5	ModPods -TDProps -Decouplers -Mk2to3 -Mk3to4	162m/s
6	ModPods -TDProps -All Parts	162m/s
7	No ModPods	162m/s

Flights 1 through 4 failed to circularize their orbit. Each of them had an average of 309m/s left on the manoeuvre node as the fuel ran out.
Flights 5 through 7 managed to circularize their orbits and had an average 162m/s dV left according to MechJebs dV readout.

The dV results change after ../GameData/ModPods/Parts/Mk3to4 was removed.

Edit: As I had more free time as expected I continued my tests.
I focussed my attention only on the three parts in the Mk3to4 folder. All tests were done under the same conditions as before except I skipped the triple tests since MechJeb has proven to be extremely consistent. Each test had a full ModPod install with only one part removed.

	Changes to ModPods	dV left/short after flight.
8	Dfuel removed	-309m/s
9	ScienceLab removed	-309m/s
10	Service removed	162m/s

It is indeed the Mk3to4 service module.

And I snapped a few screenshot at key stages in the flight:

Flight 9: To limit Q MechJeb has throttle back to roughly 50 to 55% while going 281,7m/s at 6370m. After this point MechJeb slowly throttled back up.

Flight 9: MechJeb finally throttle back up to 100% immediately followed by stage separation. 751,5m/s at 17311m.

Flight 10: MechJeb throttles back to 40 to 45% going 318,6m/s at 7776m.

Flight10: Back at full throttle. 865,1m/s at 18466m.

Flight10: Stage separation. 1005,7m/s at 20757m.

As you can see there is a clear difference in drag where there should not be any.

Edited February 10, 2018 by Tex_NL

[TiktaalikDreaming]

10 hours ago, Tex_NL said:

And now for the results of my first round of testing. I've tried to be as methodical, scientific, unbiased and reproducible as possible.

All flights have been launched from the VAB with the same 100% stock 2-stage rocket (see image below) and have all been controlled by MechJeb2 under the same conditions (also see image below).

All flights have been flown three times to average out any inconsistencies in MechJeb. (There weren't any inconsistencies. All dV results were within 0.5 m/s)
All flights started with the exact same fuel, weight, dV and TWR. The only mods involved in this test are MechJeb2-2.7.0.0, MechJeb Embedded Universal, ModuleManager.3.0.1 and ModularPods-1.0.9. The only chances between fights have been made to ModPods. The other mods all have stayed identical.

Flight #1 started with a full ModPods install. Flight #2 had only the TDProps folder removed. Each consecutively had one more sub-folder removed until none was left.

	Changes to ModPods	Average dV left/short over three flights.
1	Full ModPods	-309m/s
2	ModPods -TDProps	-309m/s
3	ModPods -TDProps -Decouplers	-309m/s
4	ModPods -TDProps -Decouplers -Mk2to3	-309m/s
5	ModPods -TDProps -Decouplers -Mk2to3 -Mk3to4	162m/s
6	ModPods -TDProps -All Parts	162m/s
7	No ModPods	162m/s

Flights 1 through 4 failed to circularize their orbit. Each of them had an average of 309m/s left on the manoeuvre node as the fuel ran out.
Flights 5 through 7 managed to circularize their orbits and had an average 162m/s dV left according to MechJebs dV readout.

The dV results change after ../GameData/ModPods/Parts/Mk3to4 was removed.

Edit: As I had more free time as expected I continued my tests.
I focussed my attention only on the three parts in the Mk3to4 folder. All tests were done under the same conditions as before except I skipped the triple tests since MechJeb has proven to be extremely consistent. Each test had a full ModPod install with only one part removed.

	Changes to ModPods	dV left/short after flight.
8	Dfuel removed	-309m/s
9	ScienceLab removed	-309m/s
10	Service removed	162m/s

It is indeed the Mk3to4 service module.

And I snapped a few screenshot at key stages in the flight:

SNIP

I should snip some of that, but it's pretty horrible trying on mobile.

Thanks for the detail. I've managed to replicate using mechjeb, and my laptop.  Dunno why it should be different. I'm seeing a smaller change in delta v, but it's reproducible, and is likely just due to different craft designs.

I'll take a close look at the service module when I get to my main PC.

 

EDIT:
And... well, that deployable antenna, not only was the antenna config pointing at the wrong module for animation, but the model doesn't actually have an antenna, deployable or otherwise.

Correcting that and I'll test on my setup here.

derpy derpy derp derp mcDerp.  

Edited February 10, 2018 by TiktaalikDreaming

[TiktaalikDreaming]

And, a fixed up cfg file.  I'll build and clean up stuff as normal for a release on spacedock

https://www.dropbox.com/s/olj5eup2eb8whrn/3to4ServModule1.cfg?dl=0

 

[HebaruSan]

51 minutes ago, TiktaalikDreaming said:

And... well, that deployable antenna, not only was the antenna config pointing at the wrong module for animation, but the model doesn't actually have an antenna, deployable or otherwise.

Any clues yet as to how that resulted in changes to the aerodynamics of other parts? Is there a related exception in the log? Maybe KSP throws an exception while loading that part, and doesn't finish processing all of the drag cubes...?

[TiktaalikDreaming]

1 minute ago, HebaruSan said:

Any clues yet as to how that resulted in changes to the aerodynamics of other parts? Is there a related exception in the log? Maybe KSP throws an exception while loading that part, and doesn't finish processing all of the drag cubes...?

I haven't even confirmed that's the fix yet, but yeah, there's no obvious link.  But it may do something weird with drag cubes, as the deployable antennas can adjust drag, so there must be some processing. Which would fail.

At last can reproduce now, so I can grab all the logs I can find.  (Loading debugstuff onto laptop soon)

[Tex_NL]

1 hour ago, TiktaalikDreaming said:

And, a fixed up cfg file.  I'll build and clean up stuff as normal for a release on spacedock

https://www.dropbox.com/s/olj5eup2eb8whrn/3to4ServModule1.cfg?dl=0

 

Downloaded and installed into my test set-up. Tomorrow I'll test it under the exact same conditions as I did before.

Good to see you've been able to reproduce the issue. I really hope you'll be able to find and fix it.
In the mean time I re-installed ModPods into my career game and simply deleted the Mk3to4 service module. It's a part I probably won't use a lot anyway.

1 hour ago, HebaruSan said:

Maybe KSP throws an exception while loading that part, and doesn't finish processing all of the drag cubes...?

I guess it must be something along those lines. I can't imagine what else could screw up unrelated parts.

[TiktaalikDreaming]

1 hour ago, Tex_NL said:

Downloaded and installed into my test set-up. Tomorrow I'll test it under the exact same conditions as I did before.

Good to see you've been able to reproduce the issue. I really hope you'll be able to find and fix it.
In the mean time I re-installed ModPods into my career game and simply deleted the Mk3to4 service module. It's a part I probably won't use a lot anyway.

I guess it must be something along those lines. I can't imagine what else could screw up unrelated parts.

Well, I've tested, and this does fix it.  I've also done a comparison between the drag cubes (adds file to list of excrements to ask for with bug reports).  And it appears that after it fails to generate a drag cube for the service module, it does keep generating further cubes.  But they're all different, sometimes significantly, from what they should be.  It's like it carries something over from the failed attempt.

[TiktaalikDreaming]

OK, there's an update out on both github and spacedock.  Spacedock as usual has DDS textures and so on.

Confirmed not to cause me an issue when used on the rig I tested with and replicated on.  Beyond that I'm not 100% sure, but having those drag cube differences is pretty close to a smoking gun.

 

 

Here's a quick diff on drag cubes;

*** 193,200 ****
      url = ModPods/Parts/Rengine/ShieldEngineAerospike-sz3/RomboidAerospikeSz3
      DRAG_CUBE
      {
!         cube = Fairing, 7.09,0.763,2.028, 7.09,0.763,2.028, 13.87,0.6861,1.049, 14.75,0.9136,0.8007, 7.095,0.7597,3.433, 7.095,0.7554,2.028, -4.768E-07,-0.4337,0, 3.872,1.347,3.872
!         cube = Clean, 6.273,0.6742,2.028, 6.273,0.6742,2.028, 13.87,0.6861,1.046, 14.75,0.9066,0.7523, 6.273,0.6706,3.433, 6.273,0.6701,2.028, -4.768E-07,-0.4097,0, 3.872,1.299,3.872
      }
  }
  PART
--- 194,201 ----
      url = ModPods/Parts/Rengine/ShieldEngineAerospike-sz3/RomboidAerospikeSz3
      DRAG_CUBE
      {
!         cube = Fairing, 4.989,0.759,2.028, 4.989,0.759,2.028, 11.72,0.7464,0.9948, 11.72,0.8501,1.086, 4.989,0.76,2.028, 4.989,0.7578,2.028, -4.768E-07,-0.4337,0, 3.872,1.347,3.872
!         cube = Clean, 3.715,0.569,2.028, 3.715,0.569,2.028, 11.72,0.7464,0.9933, 11.72,0.8292,1.04, 3.715,0.5653,2.028, 3.715,0.5723,2.028, -4.768E-07,-0.4097,0, 3.872,1.299,3.872
      }
  }
  PART

I can't really say how it gets it wrong, as drag cubes are a bit of a mystery to me, but there's clearly a significant difference.  Again, this happened to some degree or other, to every part's drag cube that loaded after the service module.  And it seems to process alphabetically, meaning Squad came after ModPods.  (serious, not serious considering renaming all my mods to prefix ZZZ in front)

 

[Tex_NL]

@TiktaalikDreaming I just ran a test with your new .cfg under the same conditions as before. My test rocket performed just as it should and reached orbit with the same 162m/s left in the tanks. I believe you've nailed it.

[TiktaalikDreaming]

12 hours ago, Tex_NL said:

@TiktaalikDreaming I just ran a test with your new .cfg under the same conditions as before. My test rocket performed just as it should and reached orbit with the same 162m/s left in the tanks. I believe you've nailed it.

Awesome. Great to have confirmation. Thanks for first highlighting this and then for sticking around to help find the solution.

[dlrk]

@TiktaalikDreaming

There are three Size 3 engines, is this intended? How do the two liquid fueled ones differ? Also what is meant by "the liquid fuel/oxidizer is unbalanced"?

Thanks

[TiktaalikDreaming]

3 hours ago, dlrk said:

@TiktaalikDreaming

There are three Size 3 engines, is this intended? How do the two liquid fueled ones differ? Also what is meant by "the liquid fuel/oxidizer is unbalanced"?

Thanks

Some dev engines from past issues may have snuck in.  I'll need to check that.

The "unbalance" is for the romboid engines. They're based on the idea put forward for the ROMBUS SSTO, and they use liquid fuel, squirted over the heat shield, instead of an ablative surface.  This way you get a fully reusable heat shield.  But it also means carying more liquid fuel than you might normally use, and specifically more than you can burn using the supplied oxidizer.

 

Jeebus.  I forget my own mod.  For the size 3, there's ablative, in either bipropellant or monopropellant forms, plus the ROMBUS style which has no ablator.  The two ablative engines look similar.

[Barar]

Hi TiktaalikDreaming,

I was using this mod and was encountering some issues with the mk3to4ScienceLab, it has the title = Science Lab Size 2 to 3 . Which is the same title as the mk2to3ScienceLab. I think it is suppose the be "Science Lab Size 3 to 4".

Also ran into a issue when I select and place the lab I get a null reference exception. When the module is also selected, several more errors appear, I see no flag, the antenna will not deploy and the texture seems to be missing. Also the two pods are identical in stats, I took a screen shot and have my output_log for the error put into a dropbox. I am running Modpod version 1.0.10 with KSP 1.31

Dropbox Link.

https://db.tt/v9f8ZR4dHI

I really enjoy the Heatshields with AeroPlug engines and Service Module size 3 to size 4.

 

[TiktaalikDreaming]

1 hour ago, Barar said:

Hi TiktaalikDreaming,

I was using this mod and was encountering some issues with the mk3to4ScienceLab, it has the title = Science Lab Size 2 to 3 . Which is the same title as the mk2to3ScienceLab. I think it is suppose the be "Science Lab Size 3 to 4".

Also ran into a issue when I select and place the lab I get a null reference exception. When the module is also selected, several more errors appear, I see no flag, the antenna will not deploy and the texture seems to be missing. Also the two pods are identical in stats, I took a screen shot and have my output_log for the error put into a dropbox. I am running Modpod version 1.0.10 with KSP 1.31

Dropbox Link.

https://db.tt/v9f8ZR4dHI

I really enjoy the Heatshields with AeroPlug engines and Service Module size 3 to size 4.

 

This thing?

Yeah, it's a work in progress.  And it used to have WIP written all over it (literally), but I was working on the textures some time ago, and then a version without the WIP texture has snuck in.  I'll label it more appropriately

[Prowler_x1]

Has anyone tested this mod in 1.4.1? Is it up to snuff?

[TiktaalikDreaming]

4 hours ago, Prowler_x1 said:

Has anyone tested this mod in 1.4.1? Is it up to snuff?

It should mostly work, but not so nicely with the mk1-3 pod as with the mk1-2.  Grumble.

I haven't had time for a proper testing of all the bits, so I haven't marked it as 1.4 yet

[b0ss]

The truncated toroidal aerospike is probably the coolest idea for a chemical rocket I've seen yet!