@sihingkk I found/was told that you have to be doing advanced or some other mode of compilation that results in a single file. When I made that change, I was able to package my Hoplon electron app.

how can I set different source-paths for one single task?

really I need to run boot-garden in build chain but with different source tree

@maxp you can call set-env! at any point in time but maybe explain what you’re trying to achieve first?

set-env! will change source-paths for all other tasks. right?

I have some thing like `

(comp
    (javac)
    (garden)
    (aot)

in build task. And I need different source path only for garden

@maxp why do you need that? 🙂

I'd like to have garden's css sources in separate tree

you mean as in separate directory? You can add multiple directories to :source-paths @maxp

when I put garden files in sources I have to include garden in runtime dependencies

but garden files really aren't 'source' thay are just 'css data'

boot-garden dependency could be in :scope "test"

@maxp: if you don't require the sources at runtime you don't need the dependency no?

I think the easiest option might be to compile garden first and then remove the garden related files using sift

when I have some file in source tree with (:require garden ) I should put it into dependencies

ant that files (garden classes, not css!) will be in uber.jar

but I dont need that at runtime

Only way around this is that the namespaces you aot compile don't require garden

No other way around this

is it really hard to change source-paths only for one particular boot task (subtask)?

Yes and that would solve nothing

but that worked with lein

There is probably some other difference here, or some difference in how aot compilation works

What namespaces have you defined your aot task to compile?

Only single main namespace?

aot all

Yeah, that is the problem

That will AOT compile garden

Or the css namespaces, which will in turn require garden

The source/resource-paths can only be modified outside of the Boot pipeline (the tasks combined with comp)

Inside this pipeline tasks access the state through fileset

Fileset is created when Boot is invoked (or in watch case, when a file is changed) and changes to source/resource-paths during the build won't affect fileset

BUT because aot task also uses fileset to fild all namespaces, you could filter unwanted files from fileset before aot task and add them back to fileset after the task

Then aot wouldn't see them

good idea!

(comp ... (with-filter (complement #(re-find #"^garden-files/" (b/tmp-path %)) (aot)))

(`with-filter` doesn't exist, but there are examples on how to implement it)

But it would be easier to just define single namespace to aot compile, unless you really need to aot compile everything to save a few seconds at app startup

I have no source files in uber jar

only compiled code

You probably have your clojure source in source-paths instead of resource-paths

yes

files in resource-paths will be included in output

why do I need sources at runtime?

If you don't use aot

In my experience Clojure apps (at least server apps) only use AOT compilation to compile single main namespace: https://github.com/Deraen/saapas/blob/master/src/clj/backend/main.clj

And the Clojure will compile other namespaces runtime

better find all compile errors at development time

AOT compilation can cause hard to debug problems with macros

And you'd anyway find compile problems when running tests, as that would probably compile all or most namespaces

Or when developing the app using repl

I use repl mostly

but I dont like put my source code in production jar

Well then, here is a with-files function which wraps a task so that is only sees part of the fileset: https://github.com/Deraen/less4clj/blob/master/build.boot#L26-L41

if you can filter the garden files using a path filter that should work

Unfortunately the temp files in fileset don't know about their original source-path, so you can't use that for filtering

thank you

I've got a macro that is used during CLJ and CLJS build. It's using an atom that holds some state which I would like shared across it. I'm led to believe that CLJ and CLJS uses two different JVM's which means that this atom has two different states. Is it possible to share the state with the atom using something like pods? Or is it better for me to just persist the state to disk at the end of Clojure JVM and pick it back up at start of CLJS JVM?

Thinking of using this to put the atom on disk. https://github.com/alandipert/enduro