ubermain is awesome btw, don't really know how it works but it does hahaha

can someone explain why I can't connect to my repl that I booted up in docker?

@jethroksy: you probably need to use the --port option when you start the docker container

to expose the port the repl server is runing on

it's started at port 1337 and exposed

you need to specify 127.0.0.1 for the client

I get another error

can ou netcat to the port?

*can you

i guess the first thing you want to try is can you connect to the repl from inside the docker container

if you can then you want to see if the port is open outside the container

not sure how netcat works

nc 127.0.0.1 1337

no returns

or telnet

I added the -u flag and typed some random stuff it says connection refused

i'd see if i could connect inside the container first

but is it not tcp

to make sure the thing is running

alright, sec

it's running

that looks good, ok

oh i see the problem i think

you need to use 0.0.0.0 when you start the server in the container

otherwise it can't expose the port

ah

so in my build.boot i need to change the host

yeah for the server

the client will need to connect to 127.0.0.1 i believe

or the ip of the nic

i don't think you can connect a socket to 0.0.0.0

web works on 0.0.0.0

really

yup

TIL a thing

i think it's a docker thing

or a linux docker thing

it doesn't use boot2docker under the hood...?

if you're on linux you don't use boot2docker

I would if i used docker-machine + virtualbo

i think there is a replacement for boot2docker for osx now

docker-machine is super cool

oh maybe that's the one

hmm i set the host option

no visible difference

i'm looking at @jaen 's example code

he used :bind

that might do it

yup that solves it!

(and the repl runs on 0.0.0.0:1337)

sweet!

i'm not sure if we need separate bind and host

just read the docs

bind is for server

not sure about the terminology here

host always sounds like a server thing to me

then again as with OSS changing it would be breaking

does ubermain pack stuff in :resource-paths too?

hmm i inspected the jar and they're inside

@micha, @jethroksy , 0.0.0.0 specified as an address while connecting is equivalent to 127.0.0.1, but it's a sort of an implementation detail that people probably shouldn't rely on. Not even sure it works in all OSs/browsers

@pesterhazy: hah, I remember us having this conversation!

ah, that was you

is there blessed way to load an initial ns when the repl starts? Like lein does with user.clj

https://github.com/boot-clj/boot/blob/master/doc/boot.task.built-in.md#repl

there's an :init-ns option, @dm3

ok, will try that

Good morning!!

Hopefully a quick question… I can now build a uberjar with boot, but how do I include the generated JS files in the uberjar?

they are missing at the moment 😞

hmm ok.. the contents of my resources/ seems to be included in my .jar file… but NOT in the resources dir.

and thus my compojure server can’t find them

@thomas: how do you serve those files? if you serve them using wrap-resource they just need to be in the jar not in any particular directory (unless you specify a prefix)

my compojure server serves some HTML that loads some JS and some CSS.

previously I had those in my resources/ dir

core/aot still broken. it looks like core/add-resource is having no effect. i inserted a line to print the compile path, so i can see that the files are indeed being compiled to a temp dir in ~/.boot/cache but they are not being copied to the target dir.

the only way i can get it to work is to drop the fileset stuff and just set the compile-path to the output classes directory.

@mobileink: have you tried looking at the fileset with boot aot show --fileset?

didn't know about that, nice. it shows the correct list of files.

then boot aot -a emits a "Compiling..." message for each of the .clj files listed by boot aot show --fileset. but no class files are produce in the output dir.

@mobileink: what does the complete task you're using look like?

just boot aot -a . do you want my build.boot?

@mobileink: try boot aot -a target

ok, that generates files in ./target, although I have :target-path #{"build"} in build.boot

it also copies the .css file that is in the fileset, which is surprising

Do you have a boot.properties file that sets BOOT_EMIT_TARGET?

as long as that file was in one of your :resource-paths thats expected

~/.boot/boot.properties has BOOT_EMIT_TARGET=no

@mobileink: the implicit syncing to target is being deprecated. use the target task's options to influence into which directory the fileset is synced

@mobileink: that's why the files have not been saved into build previously

is proper use of target documented somewhere? i thought tasks were supposed to be independent, no? how do i use target's options in the aot task, or in a task I write?

@mobileink: you don't—tasks work on/interact via the fileset, syncing the fileset to the filesystem is just another thing you can do with it

aha, so the actual task carried out by target is to configure syncing? that would make more sense - i never understood how "target" could be a task. oy. just realized i'm treating "target" as a noun - you mean it as a verb, right?

@mobileink: right, the target task syncs a fileset to a place on the user's filesystem

tasks only care about the fileset value, which is anonymous in that it doesn't correspond to a place on disk the user or tasks need to care about

does target actually carry out the sync, or just configure it?

it does it

in previous versoins of boot an internal/inaccessible thing did the sync, and it still does by default

but then we realized that was the wrong default and made a target task so it can be done explicitly

ok, i'm starting to grok to general design idea (finally). working in filesets does seem to require a certain shift in thinking.

yeah it's pretty weird congrats for hanging on 😄

thanks for writing boot! i'm pretty optimistic about it, actually - years ago i did release engineering for a multiplatform product so got very familiar with make and autoconf stuff, later looked into cmake and various other, then had to deal with gradle, leiningen, etc. i actually spend a few weeks designing my own idea of a build tool a few years back. boot is looking pretty good.

wow, thanks!

it seems like a good design idea, and i think using Clojure is the killer feature.

the key thing was lifting the problem to the clojure value level

"what if the filesystem was an immutable value"

i hope other things do this in the future

i have a sad every time i need to use other build things now 😢

once I get a little more experience I expect to pitch in with some boot documentation. do you guys have one of those O'Reilly-style books in the works? not volunteering, just think boot deserves some detailed docs and handholding for noobs.

i'm not aware of one... but yeah that would be awesome

and thanks in advance for docs help

i feel like a "cookbook" treatment could maybe serve people best atm, parts of the wiki are kind of like that

but really anything you feel like contributing is most welcome

I would recommend changing "target" to "sync" or something that more obviously reflects its function. I had no idea what it was until just now.

How does a "Troubleshooting" page for the wiki sound? I'm going to write up a brief description of the problem I just had - where's my aot output? - but there's no obvious page for it on the wiki.

that sounds excellent

hi, I have played a bit more and as mentioned above I can get my JS/CSS files in my jar file… but I seem to lose the dir structure. Is there a way the dir structure can be maintained when adding file to the (uber)jar?

I have tried :resource-paths and :directories so far.

@thomas had the same problem today

it ended up referencing files from outside the jar

i ended up with: - target/project.jar - resources/public/...

those two in some project root

@jethroksy: that is far from ideal… I would like to deploy just one artefact ideally.

understandably

are you using ubermain?

ok, might do that as a work around.

I'd love the correct solution as well

I'm using yada, I figured that out by trial and error

this my deftask:

part of it at least

the problem is unlikely a boot one

you'll need something to add the artefacts into your jvm classpath and load it from there

I wish I knew, but I don't 😕

If you add your artifact directory to the resources path, doesn't it put it in the jar and maintain the structure?

but if boot could add the files to the (uber)jar and maintain the dir structure, wouldn’t that solve the problem?

@laforge49: it doesn’t seem to do that 😞

Unexpected!

I agree.

Create an issue!

yes… I will.

the directory structure is maintained in the jar

it's just that whatever you're using to serve up the stuff isn't referring to it

@juhoteperi: should @import "../module/_xyz" work with sass4clj?

FYI I just added two new pages to the wiki: https://github.com/boot-clj/boot/wiki/Boot-Troubleshooting and https://github.com/boot-clj/boot/wiki/FAQ, with info on BOOT_EMIT_TARGET and the target task.

Please take a look at advise if you see any misinfo.

@mobileink: excellent! i added pointer to https://github.com/boot-clj/boot/wiki/Target-Directory

Regarding ^ @micha you could, but it would work only one way: either the jar works or serving it from boot works

i think that's a problem with how the jar is constructed

you should be able to have the same classpath in the jar as what you have when you serve from boot

So it is an issue?

alternatively you can use the sift task to move things around

i don't think it's a boot issue

That's what I thought too

But it's something boot should be able to universally solve

it's a feature that the files are relative to the classpath and not directly related to the files in the filesystem

i think i'm not seeing the problem

For my case I have code like (yada "resources/public")

(yada "public")

leave off the resources, no?

If I do that

Then I have to set-env :resource-paths to #{"resources/public"}

Then the jar structure would change again wouldn't it

i don't see why

:resource-paths #{"resources"}

and (yada "public")

that should work in both jar and local dev

With public inside resources?

I can't see why I didn't try that before

I'm pretty sure it didn't work for me for some reason

But yes its not fundamentally a boot issue

the idea is to provide a consistent interface to the files and whatnot

via the fileset

tasks don't know or care where the files come from

and in fact the files in the fileset are anonymous unnamed objects

Understood

associated with a resource path in the classpath

Another thing: in 3.0 are the files still going to be locked?

Files output in target

in 2.6.0-SNAPSHOT it should be fixed

Ah

can you try and let me know if it works?

I had some problems with ubermain on 2.6.0 so I downgraded

you should be able to put 2.6.0-SNAPSHOT in a boot.properties file

hm

yeah there could be a regression

I'll upgrade again in the morning and let you know

which is why it would be good to test

It was a class path thing

Couldn't find some core namespace

I'll try first thing in the morning

Boot wouldn't release the locks on my target files even if shut down gracefully within docker

oh, docker is another thing

also boot doesn't lock anything

Eh

But if I do boot dev I see the locks too

@alandipert: while we're at it, ok with you if I start adding wikipages for the predefined tasks? I've found I have to do a considerable amount of time digging around on the web to figure out how to use them, e.g. checkout (not to mention aot and target). might as well put the info on the wiki where it can be easily found.

They disappear when I stop the tasl

i didn't write any code that locks files

Odd

Anyway I'll try snapshot

how are you seeing locks?

File permissions become read only

In fact .nrepl-port is also affected

ah

that's windows

@mobileink predefined tasks have in built docs

But I'm running arch on all my machines

@mobileink eg boot repl -h

In fact most community tasks work the same way

files in the fileset are set to be read only

because if you write to them the fileset will be corrupted

and behavior from there on out is undefined

the fileset files are in an immutable content addressed store

Iirc you mentioned they were hard links

hard links to those files in the store yes

so the links must have the same permissions as the files they link to

The hard links are released when boot is terminated

they're filesystem hard links

they still point to things in the temp directories

Oh

but those temp directories are no longer referenced by anything when boot exits

and they will be deleted the next time boot runs

I see

that will leave the hard links remaining

That makes a lot more sense

I was visualizing it wrongly

In many ways I make a terrible programmer xD

@jethroksy: That's true, and once you know what you're doing boot <task> -h is very helpful. But when you're learning - well, when I was learning, at least - usage summaries are not very helpful. As I mentioned, I've had to spend a lot of time searching the web for explanations and examples, and even then I've had to go to the source to really understand. The point of wiki docs is that you have room to actually explain things.

So as you understand something that took some effort, add it to the wiki. If it is wrong, it can be corrected. If it is write, it helps the next guy.

The thing is, the topics you have difficulty with will vary from person to person. And who will take the time to carefully document something unless they were the one having trouble reasoning it out.

So there is never any point in complaining about the lack of material in the wiki. Only with complaining about inaccurate or dated material in the wiki.

And even then, once you've confirmed that the wiki was wrong, you can fix it yourself.

For myself, I have no js background and find even the simplest things in clojurescript rather difficult. So I've written a series of demo programs. https://github.com/aatree/aademos

I was just pointing out that it was there, in case he didn't know. No complaints were made

I thought this was a complaint: As I mentioned, I've had to spend a lot of time searching the web for explanations and examples, and even then I've had to go to the source to really understand. The point of wiki docs is that you have room to actually explain things.

@jethrosky: no prob,

@laforge49, @jethrosky: didn't mean it as a complaint, just an explanation of why I want stuff on the wiki

sure

I've got tons of issues like that myself.

I feel like the main turnoff for boot is that there are quite a few concepts to familiarize yourself with

When things stop working its hard to figure out what went wrong

The only reason why I can manage at all is because my needs are small and I've gotten a lot of help.

Really takes a while to ease in

IMO its not something docs would immediately solve

How many days have I lost due to the error messages coming out of clojurescript? (shudder)

Which is why micha has been such a huge help answering all our questions

Indeed!

@jethroksy: depends on the doc. really well written documentation goes a long way. unfortunately writing great doc is maybe even harder than writing great code. but we should try anyway.

If you do a bad job, some one else may fix or replace it. "That doc was really bad--even I can do better!"

@seancorfield: boot-expectations ..... I think I might like tests again :)

Glad to help steer you back to the light 😼

@martinklepsch: Re sass4clj, it's highly probable that there are some problems with relative imports currently, but if they work in other implementations they should also work with sass4clj (i.e. if they don't, it's a bug)

@micha I have made some progress on this now (having read your conversation above)...

but still not quite there…

I get this in my console: GET http://localhost:3001/js/app.out/cljs_deps.js

and that file is in: public/app.out/cljs_deps.js notice the missing js

you can move your .cljs.edn file

into a js subdirectory

is that a sift task?

i mean the source file

although really where the deps.js file is should be an implementation detail

i would just change the html so it loads from wherever the js is

but this in from the generated JS code… I have update my html all ready

strange

yes.. I am very confused… but thank you for all your help so far. it has been most welcome!!!

is there a way to tell the cljs compiler where to put its output. in the cljsbuild on lein there is.

you do that by creating a .cljs.edn file

the location of that file provides all the configuration needed

ok, I’ll have a look at that one.

because with lein the :output-dir, :output-to, :source-map, :source-map-path, and :optimizations are all dependent on each other

if you start with just a simple thing, like this

mkdir src/foo
echo '(ns foo.bar) (println "hello")' > src/foo/bar.cljs
boot -d boot-cljs -s src cljs

with no build.boot file or anything

that's the best way to start

then you can create a .cljs.edn file

and so forth

without compojure etc

ok I found it…. going to play with that tomorrow.

thank you again!!!

np! 🍻