Demo Parsing Issues with Open Fortress

[ktwrd]

Heya!

I'm currently trying out analysis-template-rust, which uses this library, and I'm having some issues w/ parsing demos for Open Fortress (the current public version, and an internal dev build on the new 64bit engine branch)

I've tried multiple demos, recorded from the server and the client, but none of them will properly parse :/

At least it fails on the same thing every time, which is Malformed utf8 while reading string when parsing the StringTables packet. Which happens with the dev build (TF2 SDK & 64bit), and the latest public build (v21, previous2021 branch)

Also here are all the demo files;

• https://res.kate.pet/upload/adb17bf0dcf6/of22_2025-04-24.dem (dev build, recorded from client)
• https://res.kate.pet/upload/8d2cb4ba985f/auto-20250423-1207-dm_bloodcovenant.dem
• https://res.kate.pet/upload/7b33b6d5508b/auto-20250423-1144-dm_rig.dem
• https://res.kate.pet/upload/a0298364502b/auto-20250422-1405-dm_byre.dem
• https://res.kate.pet/upload/024e0d1a10b4/auto-20250423-0920-duel_geoteknik.dem

FYI, there are only 4 extra string tables in Open Fortress;

networkstringtable->CreateStringTable( "ServerGamemodes", 256 );
networkstringtable->CreateStringTable( "ServerMutators", 256 );
networkstringtable->CreateStringTable( "ServerVoteSettings", 256 );
networkstringtable->CreateStringTable( "ServerVoteOptions", 256 );

If you need any more info, please let me know :3

[glyphpoch]

I'm running into a very similar issue but with stock TF2. Still investigating, and no idea if this is the same issue, but I'll leave this here for now:

Our server's got a very large mapcycle.txt file (12kb) which gets loaded into the ServerMapCycle string table. Older demos from this server only contain default maps in this string table, so this wasn't an issue before, but now the actual file started getting loaded in and is breaking our demo parser (not sure if this was a configuration change or something that changed in the last TF2 update).

Anyway, if the parser tries to load more than 8kb of extra data, then it ends up in a bad state because of this I think:

parser/src/demo/packet/stringtable.rs

Line 157 in e745137

#[size = "byte_len.saturating_mul(8)"]

saturating_mul on a u16 will clamp any value that doesn't fit into a u16 variable to u16::MAX, which will then cause the bitbuffer to cut off some data for table entries larger than that, and likely fail out with a Utf8Error when it tries to read the string table name for the next table.

[glyphpoch]

I think I've got a working patch, let's me parse demos that were previously failing - glyphpoch@4327217

I can also parse at least one of your Open Fortress demos btw: auto-20250423-1207-dm_bloodcovenant.dem. So it probably is the same issue.