#techempower-fwbm

TE-msmith: I don't know what you mean by that - are you saying that the toolset halts if any of the tests fail?

Some details needed before I know what you mean :-)

yeah

that's what I mean

I tried to execute that line above, and it died on Elixir because it failed to install correctly

we might have to address that at some point

i'm home now

much less worried about it

lol

ha

fair

I can gist you the output

yea that'd help

I suck a bit with human summaries...much easier to quickly scan a few thousand log lines :-P

by the way, is that forcibly thing working out?

can't tell until I run a full preview

Oh thta's just because of this line

You defaulted to "install_error_action" of "abort"

host

hot*

hence this crap: https://github.com/TechEmpower/FrameworkBenchma...

wait, where is that set?

ok

Well did you pass it in?

benchmark.cfg

neat

If not, then its in your config

yea...

so, I want to run this and get stdout going to out.log and stderr going to err.log

how do?

using bash? I'd send them to the same file using <command> &> your-file.log

if you send them to two files it's hard to associate them later - e..g you loose ordering

k

Once your command is runnign, use somehting like tail -f your-file.log to watch stuff happening

yup, i'm on it

just want to make sure it gets past elixir and I've kicked off a preview run

ah yea

hotness

we're winning

preview run in progress

aight; ttyt

thanks for the help

n

np

[FrameworkBenchmarks] idlewan opened pull request #1175: Update Nimrod to 0.10.0 (master...nimrod_0.10.0) http://git.io/nmdixw

https://travis-ci.org/TechEmpower/FrameworkBenc... what the heck is going on here?

I can actually see the responses (line 2844, for example) are correct, but then it fails

don't know

For some reason this fails - https://github.com/TechEmpower/FrameworkBenchma...

package 'json' is reporting that the body string is invalid json

line 2826 looks wrong too - it's not returning 96 extra bytes. Are we sure curl is actually returning the body only and not hte headers+body: https://github.com/TechEmpower/FrameworkBenchma...

unless it was changed, curl should be only returning the body

i can test in a second (setting up another piece of hardware since my main hardware server is still running the benchmarks ^_^)

jees, nickserv is slow today

can confirm: `curl -m 15 -s http://www.google.com` returns only the body and not the headers

one thing I'll point out is that cpoll seemed to be giving a body with leading/trailing whitespace

maybe the json decoder doesn't like that

gunna test

it seems cpoll isn't working for some reason: https://travis-ci.org/TechEmpower/FrameworkBenc...

(as in serving the test implementation rather than executing it)

acidx: whoa, that's weird output

is that how cpoll returns on HTTP 500?

where are you seeing this?

line 2111

it seems it's the content for frameworks/C++/cpoll_cppsp/www/plaintext

btw: this is what I'm seeing on our hardware https://gist.github.com/msmith-techempower/4371...

except look at line 2815

it's returning the correct content there

yeah. pretty weird.

2844 was what I care about

that's obviously correct and it gets a fail

TE-msmith: the underlying issue is that 2844 is showing you the header+body, whereas verification is checking what's shown in the body. From line 2111 + line 2815 those two are different things. That's whats crazy

that doesn't compute

because when I ran it just now, it worked fine

that's what that gist I linked is

Look at the travis output you posted, but look at line 2110 instead of line 2844

you will see the same craziness

here, I updated it with the console output https://gist.github.com/msmith-techempower/4371...

I see the travis craziness - I claim that it doesn't happen

so it's a travis issue?

I don't know

more likely a "doesn't work on ubuntu12" issue?

are all those compile warnings normal for cpoll?

yeah

hm, perhaps?

odd that it's clearly working 1/2the time though

(inside travis)

so

just after it compiles (all those warnings)

you should see INFO:run-ci:module /home/travis/build/TechEmpower/FrameworkBenchmarks/frameworks/C++/cpoll_cppsp/www/forcedynamic.cppsm loaded

there are a FEW (not many) error loading module /forcedynamic.cppsm: Compilation error

"error loading module /forcedynamic.cppsm: Compilation error"

it is probably no coincidence that it is trying to load `forcedynamic.cppsm` 32 times

likely running on a machine with that many cores (though, not that many available)

I may have to ping the cpoll guy

that's easier to follow

if you search for "forcedynamic.cppsm loaded"

you'll find what I'm talking about

TE-msmith: hamiltont: I am still confused by my pull request for Rust/Irion, it _looks_ to me like travis error and not error on my part. Can any of you confirm/deny ? https://travis-ci.org/TechEmpower/FrameworkBenc... https://github.com/TechEmpower/FrameworkBenchma...

I can confirm right now; sec

I'll point out that all your response bodies have double what they need

also, just confirmed - there is still a probelm

double what?

I would need the output from cargo build --release

it's missing

here's what I get from that https://gist.github.com/msmith-techempower/21b2...

oh noes, compiler errors

^

it's in YOUR code torhve!

lol

my fork, not my code :-)

Ok, then I'll try to freshen up local environment and commit fix

ah

My changes are merged upstreamed now, though, so fork not needed

(I think)

thanks for helping me, TE-msmith

sure thing

let me know when you commit again and I'll try pulling and testing again

will do!

travis is being... unrtrustworthy

they renamed stdmacro fail! 5 days ago "steveklabnik. cgaebel: fail! is now panic! as of last night"

hehe..shows you how usntable that language is

oh good lord

#include <stdio-nope-can't-touch-this.h>

the nice thing about rust is that code written in it don't even have time to get rusty. :)

>_<

so, one of those questions is "why do you hate me, travis?"

oh man, the dstat output alex-techempower coded up is insanely cool once I figured out how to plot it

You can see so many metrics about how a framewrok perormed during test run

With a hardware profile of things like max network bytes available, this could probably be used to quantify how much each resource is consumed during a framework run - would be great for quantitatively identifying which resource is the bottleneck

mmmyeah

I know some of those words

ha ha :-P

It's enough data to make any sysadmin cry with happiness?

so... yay?

sounds like yay

dear wt, why does it honestly take 20m to build you?

Very yay - just needs to be visualized and then have some code searching for common patterns

coolio