Bug Reports FEB2015c

[Thothie]

Hey, don't know if you've been told but the_wall is broken

It doesn't allow anyone to spawn, and just instakills them each time.
Thanks

Yeeah... Got that... Entity scramble issue - apparently the msarea transition is now a trigger_hurt.

I need to update the top post, really...

I will make a new bug report thread for D, when we finalize and make new installers.

 

[MS:C community]

Well, shad_palace secret room quest thingy went fine on RKS #02 this time. Crash was bad luck indeed.

This still happens though...

And now that I'm on the subject of ancient bugs (don't think these have been reported though):

• In Deralia, the NPC named "Cathain, Militia Quartermaster" (deralia/guard_barracks) appears twice. I know this is a fantasy RPG and all but... clones?
• During a certain quest, Slinker (deralia/Slinker) will direct you to Holten, the innkeeper, but the innkeeper is actually named Gerald.

 

[Thothie]

Eew... There's more than one instance deralia/guard_barracks ? Where's the other one?

 

[MS:C community]

The second one spawns at the docks (and the first one in the barracks).

Maybe one of them is a time traveler?

 

[Thothie]

Server will crash if they ever meet each other, no doubt.

Never noticed the one by the docks... Meh, tricky to fix without ripenting Deralia - putting it on the backburner.

 

[MS:C community]

The power of RipEnt has granted me access to the_wall (on my non-[FN] listenserver) and it's severely messed up.

• The trigger_once that is supposed to start the first battle doesn't work... Or, more likely, it is now a different brush-based entity... Hard to tell what though, as it's still invisible.
• The trigger_once that spawns the skeletons at the top of the staircase (before the ice reaver) is now an invisible solid.
• The breakable part of this staircase is gone.
• In the above image, I'm once again blocked by an invisible brush (presumably a trigger_once previously).
• The entire area from just beyond the spawn up to the above image is now the transition area to the_wall2.

...and that is when I gave up RipEnting stuff out of the map. Looks like a whole slew of brush-based entities had their classes changed; it's not just limited to that msarea_transiton <--> trigger_hurt mix-up.

 

[Thothie]

Yeah, I figured it wasn't, I had to recompile to fix it.

Generally, when the entity list gets off by one, ALL the entities shift over by one. Not sure how the game map and the RMF source got out of sync, but meh, nadda can do, requires a client patch now.

In anycase...

Will give it another day to see if something new major comes up, but I think we're about ready to replace our installers with a D patch (that'll fix the_wall as well - sadly, requires client update though). Judging by the top post, everything major's been fixed.

Do need to test the rmines boss fix MP though - that is, the one where he suspends monsters spawns.

 

[TheOysterHippopotami]

The power of RipEnt has granted me access to the_wall (on my non-[FN] listenserver) and it's severely messed up.

• The trigger_once that is supposed to start the first battle doesn't work... Or, more likely, it is now a different brush-based entity... Hard to tell what though, as it's still invisible.
• The trigger_once that spawns the skeletons at the top of the staircase (before the ice reaver) is now an invisible solid.
• The breakable part of this staircase is gone.
• In the above image, I'm once again blocked by an invisible brush (presumably a trigger_once previously).
• The entire area from just beyond the spawn up to the above image is now the transition area to the_wall2.

...and that is when I gave up RipEnting stuff out of the map. Looks like a whole slew of brush-based entities had their classes changed; it's not just limited to that msarea_transiton <--> trigger_hurt mix-up.

How the fuck is this even possible?

 

[Thothie]

It happens when you ripent or hlcsg -onlyents a map with an entity table that has a different number of brush entities than the BSP you are editing. It screws up the entity indexing, so things get swapped around - like msarea_transition brush, suddenly getting swapped with a trigger_hurt brush, etc. etc. Probably every brush with an entity tied to it is shifted over one or two entities over.

TIs why once you add or remove an entity brush in Hammer, the source and BSP are no longer compatible. Just, not sure how or when that happened. Recompile it completely and you're fine though - but of course, then ya gotta port the BSP to clients as well or they'll get the "map does not match server's" error.

 

[Thothie]

Alright - ~12 hours with no new reports... Any of ya see anything major or quick to fix missing on the top post's list?

edit:
(Six hours later)
Missed nothing? Really? Alrightie then...

 

[D.Drew]

Felewyn Shard doesn't seem to do anything at max charge


Tried repairing a golden axe yesterday, it caused the server to crash instantly. (tried it twice)

 

[Thothie]

Hold +use when reaching max charge, as per the changelog (albeit, spoiler sectioned). Always found it too easy to accidentally "slip" into shield-lock.

Will make a note of the G.Axe (bit tricky to test), but already uploading so...

 

[Monika's_BFFEx0256]

FN keeps flickering in and out, and The_keep just randomly crashed. Rmines also has the strangest form of lag. You could not move at all, like you were stuck with a rubber band... but mic chat was not laggy in the least. Pings spiked to 300-600... We need more servers.


EDIT: 8/8 players on hunderswamp, 1000 ping spikes...
Host_Error: UserMsg: Not Present on Client 236

People constantly getting kicked.

Maybe FN shouldn't be hosted on the same server as all the Gameservers...

 

[Thothie]

Should probably drop the max player count to 6...

...Also should probably rig up a mapper toy to cap the active mob count, maybe bypassable for boss spawns, and the like. Recently worked up a system to suspend monster spawns, which rmines is using - I think the RKS servers are using it (should suspend spawns while the boss is active).

Hope that isn't what was causing the problem - but I suspect it has more to do with the 8/8 players.

 

[Monika's_BFFEx0256]

Only other thing I could think of is not hosting FN on the same server as the game servers...

Some games only use one core of a CPU, and MS:C... being as CPU intensive as it is... may be a bit too much for even the strongest of servers if it runs by this principle. I don't really know much about how MS:C servers run or his box, but it never seemed to do this before this patch :?

 

[Thothie]

Only other thing I could think of is not hosting FN on the same server as the game servers...

Some gays only use one core of a CPU, and MS:C... being as CPU intensive as it is... may be a bit too much for even the strongest of servers if it runs by this principle. I don't really know much about how MS:C servers run or his box, but it never seemed to do this before this patch :?

Been a long time since the servers were this active.

Both the HLDS and FN are single core, but even all combined, they aren't eating much of the server's resources (usually no more than 3%), however, the HLDS has its own internal limitations, and gets confused when delays come into affect. I'll take a look and see how the threading for the processes are.

Not that it wouldn't be nice to have them on separate servers, just not an option ATM, and certainly no meaner servers available.

 

[MS:C community]

Some gays only use one core of a CPU

 

[D.Drew]

Should probably drop the max player count to 6...

oh god please no, i'll cry


pretty sure it is the amount of players though, this happened to my sven server as well... after 6 players, the pings immediately spike up to insane amounts, and it's a problem with my upload/download speed, definitely not a hardware issue

 

[Thothie]

Sadly... I've moved all the FN services to core #8... Might help a wee bit. Moving the HLDS does nothing, as it restarts so often. I'll have to ask RKS if it's possible to make TCAdmin assign specific CPU affinity. The system does auto-defer some core overhead though (anything that goes through the OS, rather than entirely internal to the HLDS) so the performance monitor does show some activity spread over all cores.

The TCP connections are insane for some reason though... Something about how Half-Life treats mobs as clients I suspect.

But yeah, this is all more software limitation than hardware... Though it's also hilarious how the HLDS takes about double the resources with every Valve patch. The 1998 processors it was built on couldn't pull more than 1FPS on the current version, I'm sure.

...and while MSC itself has been known to bloat similarly, this patch, at least, we're seeing massive optimization, at least in terms of how many CPU cycles script processing takes... Granted, certain maps are just, well, crazy insane with the number of high-resource mobs they spawn at once. (Hunderswamp among them - sucks that a player who dies and falls behind the rest, has to spawn 20 more mobs in the process of catching up with his friends.)