Use ordered id for netvars instead of shitty hash

This superscedes !980 (closed) and addresses concerns over demo compatibility.

Hold on, this doesn't address my concern regarding demo compatibility. It maintains compatibility with demos that have already been created, but writes new demos that will have the same issue outlined in my comment on the previous merge request.

Quoted for comparison:

I do not believe it is a solid approach. Specifically, it prevents replay compatibility between minor versions that add or remove cvars, whether that's versions of Lua mods or the base game in the case of Kart's netreplays, in a way that silently fails càstrophically.

You have addressed compatibility with old versions, but you have not addressed compatibility between subsequent patches/mods that introduce additional netvars.

New demo files should use plain cvar names, if it's not the case already.

...somehow I misread the diff initially, and was just about to place a new comment when you explained it yourself. Thanks for pointing it out; however, I'm confident I've read the diff properly this time and although I see where the name is read from demos, I don't see any place where it is written to them. As of 2.2.0 it used CV_SaveNetVars(), which writes the netid regardless of context, and I don't see that replaced anywhere in this diff. How is that handled?

[...] if it's not the case already.

I don't know what it does, just saying it SHOULD do that if it's not already what this branch does.

Oh, I apologise! Your comment had the vibe of having read the diff more thoroughly than I, not as a statement of what an ideal solution would be doing.

I am way too lazy to read diffs. :P

I really FORGOT

And this is why you always need to read diffs thoroughly, even if it's boring.

Thank you for being the responsible one...

added 3 commits

• 7ec0b6c7 - I meant consvar_number_of_netids to be exclusive :V
• e741a7b3 - Actually save the cvar name in demos
• d9005203 - Fuck you macro

Compare with previous version

Before I plusone this branch, what do the changes in this commit do? Does it attempt to run the block once and then exit? Why does this require the do while (0) structure rather than the previously existing curly brace context?

Only one statement can follow if, while etc. You know how you don't need a semicolon after a block? That whole block can basically be considered a single statement.

So here's an example:

if (in_demo)
   WRITESTRING(*p, cvar->name);
else

So with what the macros were before, that would turn into this:

if (in_demo)
   { ... }
   ;
else

That's basically a block followed by an empty statement. And now the else doesn't follow an if statement. It's just hanging there and that generates a compiler error too.

if (in_demo)
   do { ... } while (0);
else

In C, do while does require a semicolon to end it, so it's a single statement and everything is fine.

Bonus: You can turn those blocks into single statements by surrounding them with parenthesis too. The READ macros do this.

bruh

But it works, so plusoned.

merged

mentioned in commit 8bcc2662