it starts with an "RCC-OBJ" header .
Searching that in the EXE gives the function FUN_004546a0 which seems to load these files, including "not a valid pcode object file" errors and checks the file version against "2.17".
so it seems to solve the memory problem of loading a bunch of datafiles at start by having only one buffer, into which it loads everything? interesting idea.
so the function at 00450e00 registers some pcode debugging commands:
pc_stats
pc_listServices
pc_top
pc_dump
pc_runService
pc_spawn
pc_lookupip
pc_posttoall
pc_postto
textures seem to be 8 bits per pixel but I think they've got some per-line metadata (like Doom's sprites) and also it's one of those formats that includes mipmaps in the texture file
@foone I did the whole "Oh god, those numbers look familiar, where the hell do I know them from?" thing, before plugging it into a hex -> binary converter and immediately knowing what it was.
there's a string "VCR Connection not available" which I assume stands for something boring and video game related and not the coolest hidden feature of all time
I think Visual C++ 5 or 6 has a very annoying optimization: if you have a loop that increments a number and then at the end you only ever use that number plus one, it instead counts DOWN and NOTs the result
@foone Visual C++ 5 was the source of so much pain at work when I would port stuff from SGIs to PCs. At a later job I suggested throwing in a free copy of a newer C++ compiler with our (insanely expensive) hardware so I didn't have to jump through ver. 5's asinine hoops. I am still convinced we would have saved money.
this is a sort of thing that happens surprisingly often, thanks to null-terminated strings and the need to malloc space for the NUL terminator. and I guess starting the integer at 0xFFFFFFFF and then subtracting one for each loop and then doing a bitwise not is cheaper?
found a vaguely undocumented subcommand of a command:
the game will tell you that there's a tr_dump command (for dumping the "tractors") but it won't tell you that there's a hidden flag: "pos0", which gives different data than "position", the documented flag
so the game apparently always has networked support enabled at some level, and when you ask it how many players are in the game (by using the SV_INFO console command), it counts the players then subtracts one before telling you how many are there.
Otherwise, singleplayer games would say there are two players in the world right now. Wait, what?
now the compiler is gonna want to optimize this. The unoptimal way to do it is to just get the address of foos and then calculate &foos + sizeof(foos) * i + offset_to_member each time it needs to read foos[i].a or foos[i].e
so the obvious way to do this is to turn it into a pointer.
The pointer can just be increased by sizeof(foo) each iteration, and we don't need to do any calculations, other than doing current pointer + offset to member. That's easy and fast on x86!
but this pointer has to, well, point somewhere.
Logically it "should" point to the start of each struct foo (which also happens to be foo.a), but there's no reason it couldn't point elsewhere.
Like if the loop just needed to check the foo.e member, we could initialize the pointer at the start to &foo[0].e, then increase it by sizeof(foo) each time. Then we don't need to do any pointer math to look up the address of foo.e!
given we need to read a at offset 0, e at offset 7, and h at offset 16, you'd think it wouldn't bother using an offset pointer. just pointing it at a (aka +0) should be fine. each time we calculate e as ptr+7 and h as ptr+16, right?
my guess is that it's trying to make smaller code by using encoding variants that can only encode small offsets? and so it tries to vaguely aim the pointer in the middle of the structure, just snapping it to the nearest needed member for optimization's sake
(mostly kidding since I'm guessing this variety of VPKs is much earlier/unrelated in heritage to the Source engine variety and the packing tool used presumably isn't that sophisticated in this case)
@foone ugh, some of my worst conversations with a bad boss would have him insisting that I was not very good because I never implemented tricks like this and me explaining to him that they now had these things called optimizers and if you tried to do your own tricks you would confuse it and God knows what it would think you were trying to do and then try to optimize.
@foone Someone on the forum had an issue where they couldn't open the pdd because the keybinding is fixed to TILDE and there's no option in the menu to change it. Binding sb_togglepdd in the usercfg fixes it but it's strange why they don't show it
@foone I wonder if RCC is Ravens take on QuakeC, given the raw hex data of something like progs.dat has a bunch of function names at the start whereas scripts and default.rpc have them in the middle
@foone I don't think I am saying anything you don't know, but I would think a lot of games would benefit from their own scripting language to direct and control interactions. A story-oriented game like GTA IV would have most of the game-play logic in it's own scripting language.
@foone Check out the XBox game "Pirates of the Carribean"! cough archive.org cough
When I first unpacked it years ago, I thought they delivered its source code with the game on the disc, but it turns out they are using C as something like a scripting language that they interpret during runtime.
@foone When I wrote up my steam guide for the game I mistakenly believed it was modified Quake II because it came out in 1997 and had the same console font 🤦
Add comment