Sorry I didn't catch your message earlier. That was a bug in a previous version of the library I depend on, which has been fixed late May, Versions `2.2.0-detours11` and above should work fine. If you could send me the serveme reservation ID, that would be great too. Thanks

Related issue: https://github.com/peace-maker/DHooks2/issues/3

- Fixed the chat window not always being restored to the appropriate place

This forces default position for chat, some custom HUDs move it, practically cant do that anymore
EDIT: seems that you can with workarounds

Hats off to your dedication, that's genuinely inspiring
Do you have any set goals? Is being able to play on a server one of them?
I'd be happy to cover the costs and management of a decent global gameserver infra
I might also try my hand at getting the tf2-comp-fixes plugin features in PRs

Pazer knows how to fix this IIRC, I don't remember what the issue is anymore, but this should be recoverable

IIRC most browsers are framerate limited to 60 and have VSync turned on so you're pretty much encoding for a minuscule minority of users (those who disabled the framerate limit and vsync or those who use a media player to watch streams), moreover, you're fucking over the few that have to rely on software decoding.
Besides that, you're cramming twice the data into the same bitrate limit, making the image quality much worse.
The additional processing will most likely cause intermittent frame drops.
120fps is genuinely barely noticeable compared to 60fps.
Try it out, but there's absolutely no reason to keep it.
Sorry for buzzkill

what the fuck is wrong with you
EDIT: you're also absolutely lying about the b4nny thing

I'm not aware of any North-African TF2 communities, I would assume everyone just plays on ETF2L, you could try doing that as well.
I suggest you start a recruitment post on ETF2L (log in, go to Team Admin, Recruitment, and fill the form) and start playing some lobbies on TF2Center.
82 ping is a bit bad but it's not the worst if it's consistent, you could get used to it.

Good luck

I would like to contest this decision when it comes to the crossbow by providing some insight:

The virtual method `bool CBaseProjectile::CanCollideWithTeammates(*this)` returns the field `m_bCanCollideWithTeammates` which is initialized to false and updated to true during `CollideWithTeammatesThink`.
`CBaseProjectile::Spawn()` schedules this `Think` to execute after `gpGlobals->curtime + GetCollideWithTeammatesDelay()`, the latter returning 250ms by default.
The file concerned is `game/shared/baseprojectile.cpp`.

`CTFProjectile_HealingBolt` overrides `GetCollideWithTeammatesDelay` to always return 0ms,
however the projectile obviously doesn't collide with teammates as soon as it spawns.
I believe this behavior to not be expected, the intent seems to want it collide as soon as possible but it seems that `Think`s are only precise over a few server ticks.

One could argue that developers were aware of the effects of both methods as `CTFProjectile_GrapplingHook`, which is just below and shares its mother class with `CTFProjectile_HealingBolt`, and simply overrides `CanCollideWithTeammates` to always return false.
I don't believe this to be a fair assessment. Grappling hooks were added later on, and the only way to implement the logic to ignore teammates would have been by overriding `CanCollideWithTeammates` (or `CollideWithTeammatesThink`), possibly creating the need for a virtual method altogether.
In this case, as the behaviors are different and as the Crossbow's current implementation was probably not known to be an issue, it was overlooked and was not updated.

The file concerned is `game/server/tf/tf_projectile_arrow.cpp`.

Thank you.

Of course, there are diminishing returns, you can get a map to a pleasant state without perfecting it, in the case of Process, it did actually take only 30 minutes to switch stairs to BLOCK BULLETS and add a few transitions from displacements to brushes

This is true, therefore what I suggested wouldn't make much of a difference considering that most servers would be running TFTrue. The best would be to have TFTrue produce a whitelist file that contains the minimal amount of items possible.

I updated my post, the demo writing actually doesn't affect the server the same way that it does for CSGO because TF2 does not write the entire delay buffer at the end of a recording.
The culprit would therefore be the whitelist reloading, which I have disable on my plugin via

sm_tournament_end_ignores_whitelist 1

which is enabled by default and available on serveme.tf
EDIT: it does make it less noticeable but it's still present :(

Sorry for the bump, putting down some information I've discovered so that it can be found later, it doesn't necessarily concern players directly:

To server hosts:
- "GG Lag" is caused mainly by srcds writing the demo file to disk, which is something like ~1mbps worth of data, it happens in CS:GO as well EDIT: This is wrong, CSGO writes the rest of the delay buffer when stopping a recording, TF2 does not, this is therefore a misconception, my bad.
- "GG Lag" may also be is caused because the whitelist is re-exec'd and it has gotten huge, it happens anytime the tournament mode restarts, I will investigate to see if it can be disabled.

To map makers:
- Splash bugs are entirely mitigable by taking the time to properly clip things with `BLOCK BULLETS`, most of process can be made better by switching some `PLAYER CLIP`s to `BLOCK BULLETS` (should take no more than 20-30 minutes to fix a map)
- The fucky Wrap Assassin ball collision on Gullywash 2nd is caused by the detailing under the planks using func_lod, which should definitely be replaced by func_detail instead

To tournament admins:
- All holidays can be entirely disabled if you have SourceMod and basecommands by using `sm_cvar tf_force_holiday_off 1`
- Regardless, the Birthday ball should be entirely disabled in your tournament config via `tf_birthday_ball_chance 0`, it has a physics model and collides with stickies during rollout
- Considering most servers are able to update whitelist files on the fly, `unlisted_items_default_to` should be set to 1 and the file should only contain the items that are banned, to reduce the processing required.

Mac reminded me that stickies aren't affected as much because the physics model prevents the center from being right against the brush, still it can still happen in your average stairs

https://youtu.be/0k9ioGI6T6Q

There is a random spread and random rotation applied on pipes and stickies. It makes sense for stickies because no random would allow you to put 8 stickies down and it would look like one.

That could give decent results but it would be too hard to do via plugins, I think.

Yes, it could be tweaked so that it doesn't feel unfair when taking damage from stickies, but still work as expected when rockets are dealing damage.

That's how it currently works, if you have direct line of sight to a sticky, it will do damage to you

It's interesting to see game developers finally trying to tackle the issue, most FPS I've played don't do anything to deal with it (Diabotical, Source games (Apex included), q1/q3 and subsequent mods, (I think I recall witnessing it in Quake Champions streams), Warsaw), the exception being Reflex Arena, but I don't recall their solution being perfect (IIRC thin walls allow explosions to pass through, which should include stairs).

I think that's because of a few probable things:
- It's a difficult 3D maths problem to visualize in its entirety
- It's easy to ruin your performance budget by trying something smart (doesn't matter as much nowadays, but probably did a lot in before late 2010s)
- It has important balance and game feel implications
- The easy solution is good enough, the hard solution is a time sink (and gamedevs don't have time)

I think the change is safe in terms of balance/game feel when it comes to CS:GO, I don't see a situation where the change would make the game worse, because there's only so much control you have over a grenade, you have to take in count a lot of the physics to make it land at the right place, and you don't control the timing.

When it comes to TF2 (and possibly R6 if they have something like C4, but I don't know that game), I would see such a change as potentially extremely bad, because you can control the exact position and explosion time of stickies, allowing you to do up to 8 stickies worth of damage to someone that's completely unaware of their presence.

https://i.imgur.com/xo6gETz.png

In this image, we see how explosions currently work, the damage is only dealt if the player is in the red tinted zone. If I were to magically change TF2's code to use the new CS:GO explosion handling, the medic would take damage, and most definitely enough to die and drop.

With that in mind, there are still solutions available, the one I have in mind would be to offset the epicenter by the normal of the result of the original trace, scaled by some value, and run a new trace with the maximum length substracted with the length of the offset.

https://i.imgur.com/ckg9VuO.png

In this image: the original trace in blue, the end of this trace in orange, the normal and offset in green, the new trace in pink (max length not demonstrated).

I'm not convinced that this is a perfect solution, the more lenient you make it (via the offset length), the more you allow explosions to deal damage without the player being able to see it. There may also be degenerate cases in which the offset goes "the wrong way", but that should be easy to solve by changing how it is picked.

Hello,

This is a great tribute.

I hope you can understand that I mean in no way to be insensitive, but this unfortunately affects competitive play, as such I've taken the liberty to update my mod to allow making it not clip, and partially transparent. Here is a preview:

https://i.imgur.com/7Y1vspH.png

I'll be asking Arie to update serveme servers, for the time being, you can completely remove the statues with the command

rcon ent_remove entity_soldier_statue

EDIT: Latest version is on serveme, `rcon sm_cf fixes` will make it noclip and half-transparent like in the screenshot.

Thank you, stay safe, stay home.

It looks like you're trying to run the plugin on a Windows server, which is not supported right now, sorry

it makes you launch with -insecure, so no worries of getting VAC'd