Sorry, was out rescuing my wife. Car troubles.

might be missing something obvious.. if I want to send -D param to a task / boot repl, is this the only way: https://github.com/boot-clj/boot/wiki/JVM-Options#per-project-settings ?

@tolitius: probably; I guess that's the downside of having a single jvm

(System/setProperty "foo" "bar")

ah.. duh was trying to bend set-env! to do it

thanks Micha!

np!

tend to forget I am (`build.boot`) completely in Clojure

might be worth adding to https://github.com/boot-clj/boot/wiki/Boot-for-Leiningen-Users e.g. to point out that something like:

:profiles {:dev {:jvm-opts ["-Dparam=value" "-Xmx4g"]}}

@micha, I can PR a note about it, if that's ok?

sure that sounds great

some JVM properties are magical though

magical for pods you mean?

i mean in the JVM

since they are global per JVM

some system properties are only read when the JVM starts

and changing them later doesn't have the intended effect

right, so you can't customize them by pods (is what I meant)

right

so some properties do need to be set via BOOT_JVM_OPTIONS

makes sense, ok, I am not going to mention JVM args per say, will just give a -D example, as it is most useful (I think)

👍

oops, didn't know that you could set properties after startup

forgot it's just a wiki, so no PR is needed: https://github.com/boot-clj/boot/wiki/Boot-for-Leiningen-Users#environment-variables

Hi, I just started using boot a few hours ago and making lots of progress so far with my cljs project. (tenzing template)… but now I do need a compojure server… is there anywhere an example where I can have a look at see how it is done?

@thomas: there is boot-system which might be interesting if you're a fan of component

ok let me have a look

other than that also deraen's saapas example project uses compojure

aah ok, that might be better… I don’t thinkI need component

good morning @martinklepsch, welcome @thomas

@thomas, you can also just run random code from boot, like starting a http-kit server

you don't actually need a pre-built task for that

thank you @pesterhazy

guys i'm pretty convinced there is some real issue with boot 2.5.x wrt how it retrieves dependencies

i was talking about it 2 weeks ago here -- basically when I run boot through docker, it hangs when downloading deps that aren't in the cache

did you use the -v boot option to watch what it's doing?

it will print tons of stuff while it fetches deps

I can, you mean so we see all the aether stuff?

so you can see what's holding it up

yeah I think we realized that it wouldn't work b/c this is happening prior to when -v kicks in

I can of course try again

you're saying it doesn't happen if you load deps in your build.boot?

no, no I'm loading them through build.boot

-v kicks in before anything in build.boot is evaluated

or profile.boot as the case may be

yeah that's fine

ah ok

-v is fully in effect

so let me add the -v flag and run it inside and outside docker and see what's different

i'll report back

cool

the output when it hangs with docker is here: http://sprunge.us/AEDC

I just put the aether part

is there a chance that https://github.com/seancorfield/boot-new task will join the boot core in the future ?

that would be awesome, lein new is the only use of lein I have today

sure, i think if @seancorfield agrees sure

i think we want to explore for a while too

there are a lot of interesting possibilities in that area

what kind of other possibilities ?

various ways to form composable tempaltes

git clone is a pretty nice 90% solution

it's so much nicer to work on something runnable rather than a template that needs to be compiled and then run to be debugged

it is true that atm I am git cloning my projects and doing a replace of the old name for a new one globaly 😄

the benefit of the template thing is that it updates dependencies automatically

@yenda: I have been doing that and I don't miss lein new, and I really don't miss maintaining templates

so we don't need to keep pushing new commits to the template repo whenever anything updates

@micha: how does it do that? it queries versions?

yeah

well like the hoplon template does that

obviously it depends on how the template task is constructed

but it saves us a lot of work updating templates

which would just fall by the wayside otherwise

and when they break with some update?

is it easy to debug that?

sometimes yes, sometimes no

it's a tradeoff

yeah, probably not happening that often

hopefully

there are docs, templates, code, so many things

stuff needs to be automated or it will just go unmaintained

yeah i am a believer, just sometimes the cost of the automation hides itself in a couple of breezy first runs...

it seems like this aether process just hangs inexplicably 😕

like, the :success stage just never comes after it has downloaded almost all of the jar file

what i have observed is that this never happened with boot <=2.3.x inside/outside docker

but it does with 2.5.x, but only inside docker, which is mounting .m2/repository from the host

could be a docker change

yeah i could try reproducing this while pegging the versions of docker

if you use boot 2.3.x in the same docker container you don't see the problem?

yes

actually wait, no, i don't have boot 2.3.0 there

you can use a boot.properties file or env var

perfect, i'll try that

the binary will work with any stable boot release

but could it not be an issue with the binary itself? possibly?

i don't see how it could be

i'll try that, let me update boot.properties and run again

yeah ok

the binary doesn't do aynthing related to maven

all it does is download boot.jar, which does the maven stuff

and you get a boot.jar for each version of boot

2.3.0 worked

in the same docker container

weird

yeah... i've held off on an issue since there's a lot of moving parts, namely docker

but i think working with 2.3.0 makes it clearer maybe?

should i try with an intermediate version like 2.4.2 as well?

was there an aether version bump somewhere?

2.6.0-SNAPSHOT is probably a better thing to try

maybe it's magically fixed

ok cool

i believe in magic

it's undeniable

yeah and i don't really have a choice

how do computer work?

exactly

magic, obviously

:sad_unicorn:

no magic fix in 2.6.0-SNAPSHOT

it stalls getting pod

very weird situation

which java are you using in your image?

it smells like a URL-related timeout is being reached

using oracle java 2u55 -- one of the latest

there must be something about this on stack overflow somewhere

@micha does the new issue I just submitted --boot-reload #60--look to you like a file separator issue again?

@micha thanks for looking at it; it's an annoyance but not a showstopper... I'll hold off on an issue for now...

--I'm getting an access denied exception

@pandeiro: filing an issue would be good, maybe someone has seen this before

@micha: Ah ok, I will do it then

@laforge49: ok that looks like a boot issue now

file: "target\\main.out\\servant_demo\\core.js"

ran with 255 to be sure

with mapped files off

ugh windows

i am starting to lose hope for windows support

please don't say that!

windows users need to help fix these things

that's a shitty catch 22

i know

maybe the solution is better marketing

😄

...i.e., so windows users use boot more and thus there is more debugging power

i just don't know anything about windows development

or what the platform limitations are

maybe windows doesn't support what boot needs to do

retested with 260-SNAPSHOT and now it is working.

node went for years iirc w/o working properly on windows

--I must have used the right incantation this time!

until MS stepped in

yeah

Very happy now!

2.6.0 is doing it for me.

excellent, whew!

Yeah!

Now there are just two: https://github.com/boot-clj/boot/issues/397 https://github.com/crisptrutski/boot-cljs-test/issues/42

@pandeiro: https://keyholesoftware.com/2015/01/05/caching-for-maven-docker-builds/

you can do this with boot as well

the main idea there is doing boot ... in a layer before you build your app

Hmm. It may be that first issue can now be closed with 2.6.0-SNAPSHOT.

so it downloads maven deps

you would want to set BOOT_LOCAL_REPO in your dockerfile to something

RUN mkdir /m2
ENV BOOT_LOCAL_REPO /m2
RUN boot ...

ADD . /app
RUN boot ...

something like that

that will minimize maven downloading the world

cheers @micha -- I already have that very well optimized -- it never downloads anything twice

sweet!

but I use the docker volumes mechanism

to cache them on the host and share with the container

i wonder if that has something to do with it?

maybe some issue with java and volumes

that is a very funky part of docker...

when the classloader is loading

but the thing is, it Just Worked for >1 year already

you're on OSX right?

no no no

with OS X it does not work

that is a hopeless, windowesque nightmare that I try to take drugs to forget

ok yeah cause volumes are crazy in OSX

but with linux, it works

it, ahem, used to work

it's an enigma

it's a minor annoyance for now -- it is not even 100% reproducible

sometimes, the dependencies will finish

even in docker

perplexing

if you can test without mapped volumes that might provide some insight

ok sure, I can just disable the mount

and it will start downloading the world etc

if it goes faster without volumes then we've made progress

maybe something can be googled then lol

yep, good call, testing that now

yeah or the issue can be made more specific at least

I see that you are talking about docker, if you need specific infos about the internals "I know a guy" 😄

haah that's rgeat because we need a guy!

Looks like boot snapshot has a small problem.

in aademo/lens I reference resource but there is none. normally nothing happens but now I get this:

so there may be some backward compatibility issues

how do you reference it?

https://github.com/aatree/aademos/tree/master/lens

When you upgrade to 2.6.0-snapshot, you need to delete the bogus resource-paths in the build.boot.

i didn't see any boot code in the stack trace

I will revert and retest.

but having such fun today closing all these windows issues because of 2.6.0-SNAPSHOT!!!

Here's the full trace:

and reverting back to boot 2.5.5 fixes it.

With reload now working with boot 2.6.6-SNAPSHOT, the only issue I am left with on windows is unit testing with phantomJS!!!

does phantomjs work on windows?

Of course, I now have to turn off jetty mapped files, but that is no biggie.

yup

Here's the details: https://github.com/crisptrutski/boot-cljs-test/issues/42

I should probably retest this with the new boot as well. 😄

Hmh, what is jetty mapped files?

@micha As I'm playing around with it, I think volumes are essential to our docker-based workflow, especially to getting e.g. a live-reloading frontend env, where the UI updates when source code changes

"Jetty uses memory-mapped files to buffer static content,"?

(I'm remembering why I never pursued COPYing source into specific docker images for our app in dev -- that is a huge limitation)

jetty uses mapped io by default. On windows, a file can not be deleted if in use. So the default causes a problem. This was fixed earlier we believe, but has come up again in the new boot snapshot.

Yeah, totally, I just wasn’t sure boot-new was ready to be offered up yet… didn’t want to "presume"… https://clojurians.slack.com/archives/boot/p1454429448001656

(and "morning!")

so here's the fix for jetty, thanks as usual to micha: (serve :port 8000 :init-params {"org.eclipse.jetty.servlet.Default.useFileMappedBuffer" "false"})

Looks like the even Jetty docs mentions that memory-mapped files don't work on Windows

https://www.eclipse.org/jetty/documentation/9.2.7.v20150116/troubleshooting-locked-files-on-windows.html

Not sure if that's something that Boot can go around?

boot depends on being able to manage files in the fileset

which means deleting things when necessary

without that most things in boot wouldn't be possible

if some process is locking a file and the file changes, i'm not sure what could be done there

the operating system prevents another process from touching it to change it

actually it is strange that jetty has this problem when its running in the same jvm

i guess it's not only processes, but any open streams

maybe there is some hack that can be applied to windows

to disable that bullshit

or maybe a computed docker image is the solution

to all the windows problems

in windows if i open some system file in notepad does that mean the system can't write to it?

that seems like a huge security hole

like if i open the equivalent of /etc/passwd in notepad, nobody can change their password?

there must be some backdoor that can be done

i suppose windows will just kill notepad and then write to the file

notepad does not keep the file open except when reading or writing.

and to my knowledge, there are no backdoors to killing processes that keep files open. In extreme cases, when I can't figure it out, I log off or even reboot.

you can do it from the process explorer, supposedly

MS designed windows to make it difficult to port ux programs.

but then i think the process that has the thing open will be in an undefined state

http://www.howtogeek.com/128680/how-to-delete-move-or-rename-locked-files-in-windows/

they wanted their own software echosystem

Retch!

yeah it seems hopeless

As a long-term windows user, I'll admit to never having been a fan. And I can't stand windows path names.

i remember reading about daemons you can run on your windows machine that will delete files whenever they can

so you'd load some paths in there

and it polls them

And it really irks me when on the news they say backslash when describing a URL!

and if they ever become unlocked it deletes them

lol

hilarious

but forget all that. Have a couple on me tonight Micha! Thanks ever so much for 2.6.0-SNAPSHOT!

haha thanks

i recommend this: https://docs.docker.com/engine/installation/windows/

i bet that solves all the problems

then you can do docker run adzerk/boot-cljs:2.5.5 repl

or whatever

Do I then need to use linux?

no

?

it's just boot that will be running in a linux vm

I don't need to provide any files???

youd need to expose directories to the docker container

and expose ports from the container on your machine

windows directories??

yea

wild

@micha perhaps a link to that docker install and how to use it would be helpful for other windows users — put on boot wiki?

if it works, yes

heh

that seems like the most productive avenue to explore though

since it attacks the cause, not just some symptoms

that would fix all the other issues related to paths as well

i wonder if there is a way to package a docker image as a windows exe?

that would be sweet

especially if it could autoinstall docker if it needs to

then i could just put the docker image up there for people to download

Why not a docker image with clojure and boot pre-installed.

that's what we already have

https://hub.docker.com/r/adzerk/boot-clj/

here? https://docs.docker.com/engine/installation/windows/

if you install docker you can use it

oh

Hello all, im getting this weird message when i try to run my dev boot task:

Feb 02, 2016 10:50:03 AM org.apache.http.impl.client.DefaultRequestDirector tryConnect
INFO: I/O exception (java.net.NoRouteToHostException) caught when connecting to the target host: No route to host
Feb 02, 2016 10:50:04 AM org.apache.http.impl.client.DefaultRequestDirector tryConnect
INFO: Retrying connect

can you do boot by itself?

Clojars in down

@juhoteperi: http://status.clojars.org/incidents/pdpppjllmjff

^: @dviramontes

I followed the instructions to use the mirror, but my boot instance is still hanging with no feedback

if you have 2.6.0-SNAPSHOT you can set BOOT_CLOJARS_REPO=<other url>

alright it took its time but loaded

@micha: boot looks good, prints usage and such @juhoteperi: thanks, that would explain it right ?

hello guys, i am working on a test task, and I was wondering if you could help me to specify options from the command line for this: https://github.com/arichiardi/boot/commit/32085dc71eceebe311fabd91f1a2732da2995f7a#diff-a4c54fc631a18b940822c4cc9a9251e1R871

works fine from build.boot

(runtests :cmds '#{[test/sift-with-meta-tests]
                     [test/sift-with-meta-invert-tests]})

boot runtests --cmds "[test/sift-with-meta-tests]" --cmds "[test/sift-with-meta-invert-tests]"

throws: java.lang.ClassCastException: java.lang.Character cannot be cast to java.lang.String

probably I am not calling it right

boot.user=> (deftask foo [x xx ARG #{code} "asdf"] (with-pass-thru [x] (prn :xx xx)))
#'boot.user/foo
boot.user=> (boot "foo" "-x" "[inc]")
:xx #{[#object[clojure.core$inc 0x701a358b "clojure.core$inc@701a358b"]]}
nil

boot.user=> (boot "foo" "-x" "[inc]" "-x" "[dec]")
:xx #{[#object[clojure.core$inc 0x701a358b "clojure.core$inc@701a358b"]] [#object[clojure.core$dec 0x20bdc1da "clojure.core$dec@20bdc1da"]]}
nil

oh you're quoting it

one sec

boot.user=> (deftask foo [x xx ARG #{edn} "asdf"] (with-pass-thru [x] (prn :xx xx)))
Warning: deftask boot.user/foo was overridden
#'boot.user/foo
boot.user=> (boot "foo" "-x" "[inc]" "-x" "[dec]")
:xx #{[dec] [inc]}
nil

mmm...the arg in my case is #{[edn]} but maybe it does not make too much sense...

it is a set of seq of commands

i think that is beyond the capabilities of the simple parser thing

ok then 😄

will revert to strings

but the user can put the [] on the command line options

if they're necessary

yes, true

I will need to add an assert in the code then, ok, thanks a lot @micha!

boot.user=> (deftask foo [x xx ARG:FOO #{[edn edn]} "asdf"] (with-pass-thru [x] (prn :xx xx)))
Warning: deftask boot.user/foo was overridden
#'boot.user/foo
boot.user=> (boot "foo" "-x" "inc" "-x" "dec")
:xx #{[dec nil] [inc nil]}
nil

kind of a hack, not so good

yeah..the point was to have #{[task :opts "arg"] [task :asda :foo]}

so that you can test whatever

it works in build.boot, I guess we can live with less flexibility in the cmd line

i think you just want strings

forms are nice in build.boot, in any case everyting is converted to string passing to runboot 😄

(deftask foo [c cmds CMD ^:! [str] "asdf"] (with-pass-thru [x] (prn :xx cmds)))

like that

or something

i guess the main issue is that there is no way to do like an unparsing

like converting the other way, from clj to command line string

the parser would need to run backwards then

I will try a couple of solutions...I still need to work on deftesttask 😄

sweet!

these details will be defined later maybe during the review ok? thanks for confirming it did not work

Hello everybody, my PR for boot tests is ready for the battle (https://github.com/boot-clj/boot/pull/401):

- clone `arichiardi:sift-fix-test` (some people prefer remotes)
- make install
- cd boot/core
- boot test

I’m trying to debug a boot dev failure

where do dependencies end up with boot once downloaded?

I feel like I’m failing at google as far as finding more in-depth documentation on boot

they end up in BOOT_LOCAL_REPO, which is ~/.m2 by default

you can set it with that environment variable

@mj_langford: ^^

ty