This page is not created by, affiliated with, or supported by Slack Technologies, Inc.
2018-05-24
Channels
- # adventofcode (11)
- # architecture (12)
- # beginners (132)
- # boot (19)
- # cider (26)
- # clojure (69)
- # clojure-dusseldorf (4)
- # clojure-gamedev (1)
- # clojure-italy (46)
- # clojure-nl (4)
- # clojure-serbia (1)
- # clojure-switzerland (2)
- # clojure-uk (91)
- # clojurescript (79)
- # css (4)
- # cursive (2)
- # datomic (16)
- # docs (9)
- # duct (20)
- # editors (94)
- # fulcro (15)
- # graphql (2)
- # hoplon (1)
- # instaparse (7)
- # jobs (3)
- # lein-figwheel (3)
- # leiningen (2)
- # lumo (40)
- # mount (35)
- # off-topic (19)
- # reagent (18)
- # reitit (1)
- # shadow-cljs (123)
- # specter (7)
- # sql (5)
- # test-check (4)
- # tools-deps (38)
- # vim (20)
- # yada (9)
@kenny I just found something that may be related to your issues yesterday. In my mono-repo-of-`:local/root`-projects, if I change one of the deps.edn
files inside of a :local/root
directory, tools.deps
does not recognize the change, and uses its previously cached classpath. I have to make some minor modification to the root deps.edn
file (like, add a space somewhere) and then it invalidates the cache. So you may have configured everything correctly and clj was just using an old classpath. Can test for this by running clj -Spath
before/after changes to see when it uses the cache vs. when it generates a fresh one.
@alexmiller would it be worth creating an issue for this? I am not sure exactly where/how cache invalidation is handled
I have an issue for it
Actually looks like maybe I never filed a public issue but I have it in my internal list
Workaround is to use -Sforce
-e, --eval string Eval exprs in string; print non-nil values
is it possible to turn off the printing?when using add-lib I sometimes get errors when adding via a git SHA …. Manifest type not detected when finding deps
here is the longer stack trace
a ha!
@ghadi I did not realize that was a pre-requisite
pretty sure that is the case then
Yeah. tools.deps cannot infer third-party dependencies without some sort of manifest file
ok, cool so that’s why Maven deps are safer
for the time being at least 😉
at least I know which repos I can demo now
off the wall question … has tools.deps been tried out on graal?
Not to my knowledge
I suspect it’s bounded by network io usually
I was thinking that it probably couldn’t change the classpath
I’m guessing that an app can still expose a socket-server
(I'm assuming you mean the Graal AOT binary part, not just running Graal through JVMCI)
the repl (whether socket server or otherwise) generates classes, which is incompatible with Graal's AOT native images.
that sounds legit…
so that means - in theory at least - that you can call another language via the REPL
they have a polyglot shell so I’m guessing yes at this stage
although it seems to switch languages rather than offer FFI
hmmm this seems to be the thing
ok anyway … that’s a different channel
thanks for the help @ghadi
btw I made a small demo app with the pre-reqs in place for add-lib
I have proven it with my REPL so it might work elsewhere 🙂