Engine crashing bsp file

[frag.machine]

We are turning full circle, did you read all the posts above?

It is the user who is at harm. No, he must not need to trust the mapper or the server admin who makes his engine load a map. Again, do you trust websites with an insecure browser? I hope you keep your system and installed software up to date.

I am not ranting about the HunkAlloc "exit", as I said. If that is not exploitable, then the engine is doing almost the right thing. I say almost because the error message is cryptic to the user and to a engine code oblivious person like me sure looks like an unhandled crash. I'll try to keep in mind that HunkAlloc errors are "good" from now on.

What I do rant about (and I will not accept any excuses) are the crashes that screw up the system or allow an attacker to exploit. Those fixes are not sexy, but they need to be done in my opinion.

Also I do not consider myself a mapper. Would be great if we could avoid going on barricades on two opposite sides of the river Quake.

Yup, I read all the posts above. Have you, too ?
And no, I don't trust in insecure browsers. But I do know there's no such thing as a "secure browser in which I can rely". That's why regardless of the browser I am using, I avoid insecure content. It's nice to know that Firefox is more robust than IE, or that there's no known exploit to Opera (although I firmly believe it's just a matter of "safety by irrelevance"), but as a user I try to do my part. So yeah, regardless you like or agree with this, it's also the end user role to avoid potentially dangerous contents, because there's no way to avoid 100% of the exploits or bugs from the engine side. If you are specifically talking about better error messages, well that's something that surely can be done, but it cannot shield a dummy user from screwing things. Also, if you think people here is not doing things as they should, well, all cited engines have available source code to download, fix the problems and offer the improvements to the community, maybe in the form of a tutorial. That's how things have been working in the last 12 years, any contribution is welcome.

And finally, about the cited barricades: you're again blaming the wrong people. In more than one occasion I actively worked to eliminate such separations, but looks like people is just comfortable with the way things are. Go figure.

[mh]

Agreed that the error messages could be better, but the key point to understand is that when you see a "Quake Error" message, the engine is still in control - it's decided that things are so screwed up somewhere that it can no longer reliably continue, and is taking itself down, but it is still in control and it is engine code that has made that decision. Nothing unsafe about that.

Another interesting point I'll concede is whether or not certain fatal errors actually should be fatal at all. Sometimes you can recover reasonably gracefully - check out the errors in PF_sound for example. Sometimes it might be more appropriate to just drop the current game and dump the player to the console with some explanatory text about what just happened.

Which brings me to my next point. Error messages are one of those things (readme files are another) that the majority of users don't actually read. Even if it's a big intrusive system-modal message box, they'll still phone the helpdesk and say "...I got an error message..." - same applies to Quake error messages, and I've experience of this.

And then again, as applies so often in Quake, some behaviours might be good for a mapper, some might be good for a modder, some might be good for an engine coder, and some might be good for a player. I'd imagine that the PF_sound errors I mentioned above can be useful for catching some QC problems. Likewise when I'm coding I prefer things to crash hard and as near to the cause of the problem as possible - great for debugging!

So what is the best approach? Recover gracefully where possible seems to make sense in most cases, but what about where not possible? I'd like to hear some ideas.

[frag.machine]

Which brings me to my next point. Error messages are one of those things (readme files are another) that the majority of users don't actually read. Even if it's a big intrusive system-modal message box, they'll still phone the helpdesk and say "...I got an error message..." - same applies to Quake error messages, and I've experience of this.

^^^This.

[leileilol]

Which brings me to my next point. Error messages are one of those things (readme files are another) that the majority of users don't actually read.

Even if there is an obvious well-written clue...