Linux ET Server crash on startup with etpro 3.1.0

[[2GR]edhiltir.fr]

Hi,

I've tried to upgrade my server with etpro 3.1.0 without success (linux debian stable)

It always crash with at the same moment :

------- Game Initialization -------
gamename: etpro
gamedate: Jul 26 2004
Received signal 4, exiting...

With etpro 3.0.1, server works fine.

I've tried to start an etpro 3.1.0 dedicated server on my personal station (debian testing/unstable) with the same game config that my server, and it worked fine.

I've tried to test with an etpro 3.0.15 but links have been removed

Does the qagame.mp.i386.so of the etpro 3.1.0 has been compiled on the same computer that the etpro 3.0.1 version ? perhaps a libc issue ?

server : debian stable - kernel 2.4.25 - libc 2.2.5
station : debian testing/unstable - kernel 2.6.6 - libc 2.3.2

[Rain]

Hi,

I've tried to upgrade my server with etpro 3.1.0 without success (linux debian stable)

It always crash with at the same moment :

------- Game Initialization -------
gamename: etpro
gamedate: Jul 26 2004
Received signal 4, exiting...

What kind of processor do you have?

[[2GR]edhiltir.fr]

It's only a poor old K6-2 500

So, with your question, and a rapid google search "Signal 4" means "illegal instruction". Does the .so has been compiled for i386 or a more recent arch ?

PS: It's a server for test, not for public or war use, so dont spend to much time on this

[bani]

compiler optimizations for etpro were recently changed (by rain!) to target pentium or better cpu. k6 is a little too old for that.

[[2GR]edhiltir.fr]

Thanks for the answer.

My poor lonesone server wont be able to have new ETpro version anymore

[Rain]

compiler optimizations for etpro were recently changed (by rain!) to target pentium or better cpu. k6 is a little too old for that.

They always targeted that. gcc just actually listened this time.

[faltytower]

Just needing clarification, this optimization also affects desktop/workstation running on old processor too, correct??

[gotenks]

this should only affect servers, so if your desktop runs a server then yes, but shouldn't touch your clients

[ReyalP]

this should only affect servers, so if your desktop runs a server then yes, but shouldn't touch your clients

Eh ? assuming cgame is compiled witht the same options, it will likely cause the came error.

Mind you, if you cant find a pentium2 or better system for the price of carrying it away, you aren't looking very hard. While a k6-2 500 might serve a couple of slots with out too much lag, it would make a really awful client.

[Rain]

* Intel® Pentium® III 600Mhz processor or AMD equivalent

The Linux binaries (even the dedicated server) have always been compiled with the target architecture set to 'Pentium Pro' (which is basically a P2, sans MMX), although the optimizer apparently never saw fit to actually use any ppro-specific instructions until just now.

I'm not at all surprised that people are still using older systems for ET, though… especially since I've seen players connect with the following cards:
1 3Dfx/Voodoo3 (tm)/2 TMUs/16 MB SDRAM/KNIICD (Jul 30 1999)
1 3Dfx/Voodoo5 (tm) SLI/2 TMUs/32 MB SDRAM/3DNow!/ICD (Nov 2 2000)
2 RAGE 128 Pro C21 AGP 4x x86/SSE
479 Intel Brookdale-G
(full list)

Not sure what we'll do in regards to this, though… We might fiddle with the optimizations again, although I don't know whether any of us has a K6 to test on anymore.