Running 42 in background with no GUI causes IPC to NOT be loaded

[PeteBlanchard]

Was doing a test running 42 in the background (NOS3 build) and ran into a weird issue, was wondering if you might have an idea as to what is happening... If I start the simulation with the GUI set to false via a command like this, it works fine:

cd /opt/42
./42 <sim folder>

However, if I then try to run in the background (and save the log) it stops after it reads the Inp_Cmd.txt, but fails to process the IPS or step through the simulation, it just hangs:

./42 <sim folder> | tee /tmp/42.log

Produces:

Material material_3 is not in Matl structure
Setting material to default
0.0 SC[0] qrl = [0.0  0.0  0.0  1.0]
Reached CmdScript EOF at Time = 0.000000

Looking at the 42exec.c, it seems like it should be fine...after processing the command interpreter, it checks whether sockets are enabled ('_ENABLESOCKETS'; they are, but don't see the 'server listening'), then it checks '_USEGUI' (it is), but the 'FALSE' should set 'GLEnable' to false, so it should drop into the 'SimStep' processing...actually, those checks are at the compiler level... so they aren't even being done at runtime...

Thanks in advance.

[PeteBlanchard]

update, not just with 'tee'...anything that redirects input seems to cause this...e.g.

./42 <sim folder> 2>&1 > /tmp/42.log

If output is NOT redirected, it works as expected...

[ericstoneking]

Hi Pete,

I think you're being tripped up by a poorly-documented feature. 42 accepts two command-line arguments: ./42 InOutPath ModelPath

By default, InOutPath = InOut and ModelPath = Model. You've seen cases where you set InOutPath to whatever you want. The need for multiple ModelPaths hasn't really ever come up.

So try this: ./42 Model 2>&1 > /tmp/42.log

Regards, -Eric

[ericstoneking]

Oops. Correction to my previous comment:

./42 YourSimCase Model 2>&1 > /tmp/42.log

[PeteBlanchard]

That seems reasonable, but I get the same result (after changing), stops after executing the command script and before IPC.

[ericstoneking]

Hi Pete,

So I tried it on my installation (which I should have done in the first place). These commands all worked fine with or without graphics:

42 2>&1 | tee log.42 42 InOut 2>&1 | tee log.42 42 InOut Model 2>&1 | tee log.42

I don't have any bright ideas about what's causing your issue. This is about where I would "make clean" and break out the debugger.

Regards, -Eric

[PeteBlanchard]

Bizzare, I can't get Visual Studio Code to actually stop at a breakpoint when running...but it appears to start processing the IPC and get's stuck on this line (from Kit/iokit.c, line 159):

      if (bind(init_sockfd,(struct sockaddr *) &Server,sizeof(Server)) < 0) {

because if I kill the process (after awhile of being hung) it will print out:

Error on binding server socket.

But no idea why that is occurring when doing re-direction with no GUI...

Thanks for the help, Eric...guess not much more to do from your end.

[PeteBlanchard]

One more thing...I forgot to mention that I was starting 42 in the background...

./42 <sim> 2>&1 > /tmp/42.log &

I am running under NOS3 environment, but I wouldn't think that would be an issue (though it is a slightly older version of 42, not master)

[PeteBlanchard]

Never figured out what was going on with this...may have been VM issue...did get it working though, your code was fine.