#cuberite

?

I'm using the exact same compile script on my old buildserver and the build works 0.0

only change is the debian verison

I just removed the check if it is able to find %VS120COMNTOOLS%

No idea. Try turning on verbose makefiles and looking at the mbedtls cflags.

how might I do that

any idea how to test without filling jenkins with failed builds?

no idea. The failed builds don't really affect anything, just mess up the history, anyway.

I opend %VS120COMNTOOLS% with the explorrer and it worked

so, I removed the check if its there

I'll try reinstalling the windows sdk

hm, is there a need for a server with a bit of storage and high speed connection?

@bearbin, whats the traffic like for your build server?

Because I did some rough estimates and put them into the S3 price calculator, and it says 10 cents a month, if we use travis's build artifacts system.

For hosting the downloads.

Not 10 cents 91 cents, because amazon hides things.

how much traffic do you produce?

Thats for an estimate of 1000 downloads a mounth.

Even just moving over the x64 Linux builds to Travis/S3 would reduce the load on bearbins server.

hm, how complicated is that? I could help maybe

I just want to make sure of the estimates first, so I'd like to know from bearbin.

k

We'd need to set up an AWS account, add some lines to our .travis.yml, then point the build links at S3 rather than bearbins build server.

We'd probably want to add settings to S3 to automaticly retire builds once they reach a certian age, to avoid storing too much data.

hello SphinxC0re

Hi LO1ZB

do you know stuff and thinks about msbuild SphinxC0re?

msbuild?

thats a no?

yes :P

ok :<

[MCServer] worktycho closed pull request #2034: Moved BroadcastChunkData over to cBroadcaster (master...BroadcastRefactor2) http://git.io/vTv8e

the problem is, that I'm getting that error when I try to add my pc as slave to travis

"Cannot determine the location of the VS Common Tools folder."

What user is Jenkins running as>

?

It may be using different env vars from your desktop session

it runs as a windoes servise

windows service*

That might have something to do with it.

Go into the edit system enviromental variables window.

ok sec

and check the vars there

VS120COMNTOOLS = C:\Program Files (x86)\Microsoft Visual Studio 12.0\Common7\Tools\

And thats in system env vars. not user env vars.

yes

Then the only thing I can think of is WoW.

Is Jenkins a 32 bit process?

running from the start menu does work

java

Might be permissions.

thats a java app

So its probably not WoW. program files shouldn't have permissions issues.

I can't think of anything.

the app is inside C:\Jenkins

Whats the permissions on that?

Is it owned by your account?

open to everyone

cmake works

Write to everyone?

yes

Can't think of anything else.

thats a problem with the configuration of visual c

by default the start menu leads to C:\Program Files (x86)\Microsoft Visual Studio 11.0\Common7\Tools\

I dono have visual studio 11 installed

I did some modifications to vcvars32.bat but I cannot test if it works atm

We really need to reduce dependecies on cWorld

Changing ChunkSender.h is aan indirect dependecy of everything through cWorld

I agree

I'm doing what I can, but everyone needs to commit to it.

Fixing the god objects is far more important than the name chnange in my opinion.

cWorld, cChunk, cChunkMap, cClientHandle are the worst, in that order.

is the forum only 4 me slow?

I'm not seeing anything.

it sometimes takes 10+ sec to load

going down the rabbit hole.

to bed?

No, Fixing one thing involves fixing another first.

Which leads to another issue which leads to another.

In this cas, Fixing ChunkSending, has led me to memory ownership of cLightingThread Callbacks.

sounds complicated. o.o

I'm sure it isn't for tycho_

LightingThread required that the callback was owned by the caller.

I'm simplifing it so that LightingThread owns it.

btw, tycho_ do you have an idea how to fix this issue: http://forum.mc-server.org/showthread.php?tid=1...

all I know is that it generates a few chunk befor it crashed

if that helps

I already knew that :)

I'm not sure, which verify is it hitting?

I understand the lighting stuff.

But those multithreaded bugs usually involve rediculus interactions.

The one I had with entitys not getting destroyed took several days to fix, with a reliable way of generating them.

LightningThread.cpp line 345

Just a few quick checks.

You're sure the chunks are loaded?

I'm using the HOOK_CHUNK_GENERATING. I don't know if they are loaded. I don't think so. But I should be able to change the generating chunk. Also I should be able to query chunks fo regenerating

Are you triggering the lighting you're self?

No. why should I? I don't even know how this is possible

Ok, then its probably curruption, in which case the problem is much earlier than the verify

Have you tried running under valgrind?

No

Valgrind searches for memory leaks

Valgrind serches for memory leaks and memory curruption.

Okay

I'll try

The only two tools when fighting this sort of bugs are valgrind, and adding lots more asserts.

the server still doesn't compile on my machine though

gcc 5.1

PR 2153 has the fix, but theres another issue I'm not sure about.

build on 2153 failde

Yas, theres something else in the update thats interfering

We might have to update SQLite

But as a temporary fix, if you delete everything below line 46 in lib/SQLiteCPP/include/SQLiteCpp/Exception.h

Then it will compile in gcc 5

I dono know if it is important, it only crashed in release mode, not in debug mode

[MCServer] tigerw pushed 1 new commit to comparators: http://git.io/vkuGm

MCServer/comparators acfb4d0 Tiger Wang: Test

Then its allmost certianly curruption.

the visual studio debugger also gives no information besides the stacktrace

Curroption. Are you on Linux or Windows?

I'm on windows

Linux

Its different options for Linux and WIndows

has "_Reserved = 3" any meaning?

On Windows, if you go into setflags and add /GS on line 63 in generates binaries with more stack curruption checks

@LO1ZB What context? Its UB in C and C++

visual studio's auto and local window when debugging shows that

I have no Idea.

ok

where is that setflags?

SetFlags.cmake in the rrot dir

@SphinxC0re on Linux if you go to line 109 and add -fstack-protector-all -fsanitize=address it adds similar protections.

I'm re-running cmake....

compiling

I'm not making any promoses about the flag, just that it can help detect some forms of memory curroption.

ok

wait.. 312 chunks/s o.o

I run it to create the setting.ini

Debug mode is slow.