Why was replay format changed?

[imbaczek]

is it really that hard to check for another [GAME] after the first script? i can't understand what's the fuss about.

[TradeMark]

Yeah it isnt.

I was thinking that you could include first then player own script.txt and then the host one, that way i could get the flags/ranks/watsoever without crying when i play on crappy autohosts, and it wouldnt affect the syncing at all, since my script.txt isnt used in the game.

Actually that kind of stuff would be possible (including lobby generated script (ie. the one Auswaschbar quoted) and then TradeMark concvincing all lobby devs to still include this player information. (Since it won't be required for Spring anymore.)

Well, lobby devs don't have to include that; if both scripts are include Trademark can parse the first for myplayername and get the rest of the info from the real script.

Exactly!

[imbaczek]

I was thinking that you could include first then player own script.txt and then the host one

i'm +1 on that, actually. no point in storing redundant data in the demo.

[Tobi]

At least include a (client)scriptOffset field in the header then, scanning for the second occurrence of [GAME] (second, because client has the first one) and counting { and } seems like way too hacky to be the 'official' way to access the thing.

(I can think of multiple normally harmless changes to a lobby that would brick this search: using case sensitive search and lobby writing [game], which is fine for Spring, in the script; lobby putting some comment in the script that happens to include '[GAME]', etc.)

(I should have made everything with offsets in the header in the first place..)

[lurker]

No need to store two copies, though; you can have the offset point into the demostream.

[MelTraX]

I don't really care one way or the other as long as everything needed to access the scripts is in the header.

Just wanted to point out that the player info stuff is also redundant and I don't think a few kilobytes hurt anyone if it makes it easier to get the info.

[Tobi]

No need to store two copies, though; you can have the offset point into the demostream.

Yes, storing two copies was only a workaround because the format requires the script to start directly after the header. (At least that's how I see it.)

[lurker]

Speaking of scripts, is there any reason to limit them to 32KB as is currently done?

[Tobi]

Where is that limit? (Never seen it.)

[lurker]

The fix to make it not crash at 4KB was to up the buffer to 32KB. But there are valid reasons to want to store a lot of data in the start script.

[MelTraX]

Auswaschbar told me it could be 2^32 bits now. Didn't check it though.

Yes, storing two copies was only a workaround because the format requires the script to start directly after the header. (At least that's how I see it.)

I suggested it because that way the format wouldn't need to change. But I don't really care myself whether it changes.

[lurker]

Well the commit labeled "increase the size of the fragnment-buffer to fix crash with long script.txt" merely changes three instances of 4096 to 32768, and the two commits related to outputting the startscript don't seem to change its ability to be larger.

[imbaczek]

32 kilobytes should be enough for everybody.

oh, and don't quote me on that.

[Tobi]

The fix to make it not crash at 4KB was to up the buffer to 32KB. But there are valid reasons to want to store a lot of data in the start script.

Hmm ok. I suppose the reason is that it was easiest to make it this way (and fastest).

But I suppose it wouldn't be that hard to make buf a std::vector<unsigned char> (with sane default size, e.g. this 4K) and change the asserts to conditions that double the vector size when necessary.

If I remember when I get home I might do that.

Auswaschbar told me it could be 2^32 bits now. Didn't check it though.

Hmm, from the code in github the limit still seems to be at 32K bytes max per network message.

[Auswaschbar]

But I suppose it wouldn't be that hard to make buf a std::vector<unsigned char> (with sane default size, e.g. this 4K) and change the asserts to conditions that double the vector size when necessary.

First I wanted to do it with variable size. This should be easy because you know the size you need in the first place, so no need to increase it even. But I then got occupied by rl work, and just made it to 32kb.

[TradeMark]

Why couldnt you put my script.txt first, and then host script.txt ?

It sucks now when its 2 identical copies for nothing.

It tells me every time i am the host, MyPlayerName=Soloman; bullshit, im not soloman.