search

orbitersim / orbiter

Open-source repository of Orbiter Space Flight Simulator
MIT License
1.59k stars 214 forks source link

Debugging Orbiter startup issues #382

Open GLS-SSV opened 1 year ago

GLS-SSV commented 1 year ago

As some might have noticed, recently new users show up in Orbiter-Forum saying that a clean Orbiter installation doesn't start or crashes, and most times the log contains no info about what is happening. Without solutions, the users understandably give up, which is always bad, but especially so for a small community such as ours.

I think years ago, Orbiter had a small test app to confirm the system and installation were ready.... maybe something like that could be added, or even a (config file only?) parameter could be added to generate biblical levels of logging, or some extra tests, during the startup.

GLS-SSV commented 1 year ago

Threads of recent issues: https://www.orbiter-forum.com/threads/orbiter-crashing-upon-scenario-launch.41285/ https://www.orbiter-forum.com/threads/orbiter-not-working-on-laptop.41290/

daves-ssu commented 9 months ago

Add me a supporter of this, as I have an issue with my KSC Upgrade Project refusing to load. Gone over the config file several times and can't find anything amiss. Absolutely nothing worthwhile in either D3D9Client or the Orbiter log even with Verbose Output enabled and Debug Level set to 4.

In my mind, Orbiter's logging of fatal errors that result in CTDs is beyond atrocious and it has been pretty much since the beginning. It seems to me that only some things are actually coded to output something in the log file while some things are not for whatever reason. This is a golden chance to rectify this.

n7275 commented 9 months ago

I certianly would agree that this is needed. I think the hard part is obviously the "how".

It's very hard to look at the new user perspective when you have 20 years of experience.

There are plenty of reasons that Orbiter legitimately wouldn't be able to load, but they should be reported to the user.

Do we have ideas on how to tackle this?

On Fri, Nov 17, 2023, 07:16 daves-ssu @.***> wrote:

Add me a supporter of this, as I have an issue with my KSC Upgrade Project refusing to load. Gone over the config file several times and can't find anything amiss. Absolutely nothing worthwhile in either D3D9Client or the Orbiter log even with Verbose Output enabled and Debug Level set to 4.

In my mind, Orbiter's logging of fatal errors that result in CTDs is beyond atrocious and it has been pretty much since the beginning. It seems to me that only some things are actually coded to output something in the log file while some things are not for whatever reason. This is a golden chance to rectify this.

— Reply to this email directly, view it on GitHub https://github.com/orbitersim/orbiter/issues/382#issuecomment-1816306439, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAAYP37NJF5UXRMKHREV4MTYE5ITXAVCNFSM6AAAAAA3OGRJFWVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTQMJWGMYDMNBTHE . You are receiving this because you are subscribed to this thread.Message ID: @.***>

GLS-SSV commented 8 months ago

Another thread, this one with a positive outcome: https://www.orbiter-forum.com/threads/d3d9-client-not-working-on-widnows-11.41496

GLS-SSV commented 8 months ago

Do we have ideas on how to tackle this?

I think we can break the errors into 2 classes: CTD and "this isn't right, but we can continue". The first class would be "can't read files", "can't get graphics handle", "missing vessel", etc, the errors that prevent the sim from running. These would be imediate CTD and error message window (and if possible, log entry), with some indication of what failed. The second class would be "missing texture" or "wrong param in some config file", things that just degrade the sim. These would result in immediate log entry, and checking a variable, so that at the end of the sim, a error message window can be shown, so these errors are not missed (I think something like this exists already).

The trick here is: everything that can fail needs to be checked, and logged when in error.

Xyon commented 8 months ago

I do feel, and have felt for some while, that error handling is more or less absent in the sim. Writing to the log file seems hit and miss; there are too many occasions where the log has some startup stuff in there and then nothing, having crashed out to the desktop.

What I don't know currently for having not looked at all into it is how feasible it is to hook into every time the simulation is closing, regardless of reason, and check if an error happened, so we can both write to the log and present to the user a message to say "the error happened because of this". Does it make sense to most users if we say "couldn't get graphics handle"? I'm not sure it does.

schnepe2 commented 7 months ago

I once had made a little App, that did some checks. But it is long lost and I didn't keep it or kept te sources (reasons later) It looked something like this: OrbiterCheck

I "deployed" it to some co-workers (just 5, but hey) to test it and the result was something like this:

So after a few iterations I found it not worth the effort (lots of documentation is needed to give "proper" instructions). And guessing all the possible mis-configurations is another thing - ask someone to intentionally make a stupid install just to check that, never was popular 😜

Never the less, I totally support the idea and intention!