Alright, it's a small data sample, but comparing 00:00 - 16:00 PYST yesterday and today, caching has reduced the average server load by 50%. More importantly, the arena used to spend 9.5% of the time at load 1 or higher, but just 4.2% of the time with caching.

So load 1 is when performance noticeably drops?

A load higher than 1 means that a process is waiting for resources, which is essentially wasted time.

@Dennis what wrapper do you use to execute J code?

Nothing fancy. github.com/TryItOnline/arena-server/blob/master/wrappers/j

.code.tio is the code?

Yes.

ohh. I was trying to execute J code with jconsole.cmd. haha, cool

@Dennis Can you add Parenthetic?

@DLosc tio.run/nexus/parenthetic#rZFdEsAwBITf9yTc/…

That language is kinda insane.

Yep :D Thanks!

@Dennis Is it TC?

@Dennis Fedora still has Python 3.5?

@Pavel No clue.

@Pavel I'll look into it.

@betseg Fedora 24 does. 25 is already out though.

@Dennis what about "Hello, World!" ;)

Not even the slightest clue. :P

@Dennis one of the slowest things in the scripts for arena setup is compiling pari gp

@Dennis did you have to compile it or is there a faster way?

I don't recall. I'll check.

@Dennis can you check if I understood something correctly

echo ${1%/}/*[^~,]

In this expression ${1%/} takes the first parameter on the command line and removes trailing slash

so /root/' becomes /root`

then / adds the slash

then we get all files/folders in the current folder

but only those that do not end with ~ or ,

is this right?

Yes, that's correct.

If you don't need canonical filenames, $1/*[^~,] will work as well.

It will exclude dotfiles though.

@AndrewSavinykh I've compiled it. I don't remember why though. It's in the repos.

@AndrewSavinykh Here's Hello World for Taxi. tio.run/nexus/…

And for Templates Considered Harmful. tio.run/nexus/…

@Dennis no it is not, I surely tried it before compiling

or is it?

@Dennis where did you pull those from (as in pulling out of a hat)

Taxi was used in a Hello World anser, Templates was mentioned in TNB.

@Dennis, could you please point me at the package you mentioned for pari/gp

When I tried to use TCH, I got this

Something bad happened
Please check your browser's console for further infomation.

Only one line in the console though. Unknown property ‘column-width’. Declaration dropped.

@Dennis ^

@feersum you might want to give Dennis the link that reproduces that

@feersum What code were you executing? Hello World works for me.

Some code that results in a compile error by exceeding the max template depth.

Ap<


Fun<Ap<Fun<If<Eq<A<2>,T>,A<1>,And<Eq<Ap<Fun<If<A<1>,Ap<A<0>,Rem<A<2>,A<1>>,A<1>>,A<2>>>,A<1,1>,A<2>>,T>,Add<Ap<Fun<Rem<If<A<1>,Mul<A<2,1>,Ap<A<0>,Sub<A<1>,T>>>,T>,A<1,2>>>,A<1>>,T>>,Ap<A<0>,Add<A<1>,T>,A<1,1>>,Ap<A<0>,A<1>,Sub<A<2>,T>>>>,T,A<1>>>





, I<10> >

It's from "Compute the Carmichael function", but I changed one function to see what errors it would generate.

@feersum tio.run/nexus/…

I'm getting timeout

not the error you mentioned

@Mego so?

of course it's cached I just ran it!

@feersum That sounds like tio.run. It doesn't have any error handling yet.

Oh, I didn't understand I'm not supposde to use that address.

tio.run/nexus dutifully reports that the time limit was exceeded.

@Dennis so right now if I repeatedly in quick succession click on the run button the script that times out, will the server crash because of getting out or resources?

@AndrewSavinykh There are some restrictions in place and it shouldn't actually crash, but other submissions running at the same time would time out as well. That's pretty much the scenario I was talking about the other day.

@AndrewSavinykh dnf provides /usr/bin/gp reports pari-gp-2.7.5-2.fc24.x86_64 : PARI calculator.

@Mego Cache hits with an error automatically run the code again to prevent cache corruption.

@Dennis yep, that's why I'm asking

@Dennis so are you saying that time out should not be cached? because currently it is, at the link me and Mego were discussing

My mistake. It will run again if there's an error, not if there's a warning.

Errors reported by the frontend, that is. Empty response, language not found, etc.

@Dennis I was wondering how did you come up with the idea of separating parts of the response with random 16 characters. I have not seen this anywhere before (mime multi-part messages are similar but not quite the same)

I don't remember. I've been using that for a long time.

@Dennis just to keep you in the loop, what I'm trying to do now is rewrite the arena setup script so that it can

@Dennis Found a typo. The TIO site says Server-side permalinks, which provide shorter URLs and can be edited with updating the permalink. I think you want ...without updating the permalink

a) reboot and resume on restart after selinux labeling b) log all the output from different sub scripts in a folder and c) time indiuvidual subscripts and report on the biggest time hogs

since I'm not very proficient with bash it's going rather slowly, but it's fun

@MistahFiggins Right, fixing. That's for catching that.

Welcome

@Dennis I'm also torn about the model I chose where all the repo deps are in a single script and all the other in their own files. Presumably having each in their own files would be more logical, but I have no desire to findout dependency for each and every of 200 languages. that is I have not easy/fast way to find out if a language dependency is satisfied because of some other language dependency if you see what I mean

In this regard my intention was that for the languages that keep updating you could use the script to update them

so that eventually authors can get a button "update my language from the github or whereever"

but then you need to classify all the languages on these that come from github, those that come from a url without version (latest) those that come from a url with version (how do I get the latest then) those that come from repos, and those that are never going to be updated

Yeah, that would be useful. I don't mind pulling them manually, but that comes with unnecessary lag when I'm asleep or on the road.

So these are the things I'm mulling over in my head

Fortunately, caching seems to have bought me some time, so I can get more things done before setting up more arenas. The stats for the last 24 hours look rather nice.

Mean: 0.20384722222222226
Median: 0.09
Maximum: 3.67
Overloaded: 0.029166666666666667

Load is higher than 1 only 3% of the time.

And the cache isn't even close to full yet.

@AndrewSavinykh Btw, I updated the server repos to include the three new languages.

@Dennis appreciate that

@Dennis did you get all of these with uptime?

uptime and a little parser script.

@Dennis what does it parse? httpd logs?

No, I'm running uptime in a loop and pipe the results to a file.