nice! thanks to @ales.najmann for help 🙂

Yes, thank you both!

I commented out what looked like what was leading to that:

#_(testing "config"
    (assert-submaps
     '({:file "corpus/deprecated_var.clj",
        :row 10,
        :col 1,
        :level :warning,
        :message "#'foo.foo/deprecated-fn is deprecated"})

Unexpected error. Please report an issue.
Unexpected error. Please report an issue.
Unexpected error. Please report an issue.
Unexpected error. Please report an issue.

did you see this?

ah then those might be caused by something else... hm

in main.clj this line is printed. maybe you can include (.getMessage e) in the println

i can try that -- though in my main.clj there is something about printing the stacktrace. is there some reason i'm not seeing any info about the stacktrace?

yes, because that text is going to *err*

and normal println is going to stdout

which is not consistent in this case, but stderr is not visible I think in the tests (which could be a nice for debugging)

thanks for the explanation -- build with patch complete, waiting for test output 🙂

lein test clj-kondo.main-test
Unexpected error. Please report an issue. clojure.lang.Symbol cannot be cast to java.lang.String
Unexpected error. Please report an issue. clojure.lang.Symbol cannot be cast to java.lang.String
Unexpected error. Please report an issue. Unsupported character: \\.specs$
Unexpected error. Please report an issue. Unsupported character: \\.specs$

aha... interesting. maybe you can print the entire stacktrace to stdout? which is an argument to .printStackTrace

sounds good, just need to figure out the details of how -- will bug you if i have trouble 🙂

Probably just (.printStackTrace e System/out)

i did something different -- but may be that would have been enough. too much output so adding to issue.

github seems to not like my comment 😞

https://pastebin.com/KwdEZHSS

i guess they are still having issues

I forgot to compile with clojure 1.10.1 which I do in linux.

maybe it has to do with that, but clj-kondo also should be able to run with clojure 1.9.0, so I don't know

i'm puzzled as to why the test framework doesn't explain better which test(s) it's experiencing these issues with.

maybe it has to do with how I process the result from the shell output

may be so -- but i feel like blaming the test framework 🙂

haha

ok, i guess i'll change compile.bat to include: call lein with-profiles +clojure-1.10.1 do clean, uberjar then?

I already pushed it now

great

i allocated another core to my vm to cope with slowness..it's gradually eating away at my real computing resources 🙂

curious...`lein test` leads to this: https://pastebin.com/7a9GwMqj

I found this out:

Could not parse config! Unsupported character: \\.* "{:linters {:unused-binding {:level :off}, :unresolved-symbol {:level :off}, :refer-all {:level :off}, :type-mismatch {:level :off}, :deprecated-var {:exclude #:foo.foo{deprecated-fn {:namespaces [foo.bar bar\\\\.*], :defs [foo.baz/allowed foo\\\\.baz/ign\\\\.*]}}}}}"

> Caused by: java.lang.ClassNotFoundException: clojure.core$with_meta__5420 That's usually caused by stale target files

lein clean solves it

(already added it)

somehow the backslashes get doubled

https://stackoverflow.com/questions/34402429/processbuilder-does-something-to-my-command-line-arguments-on-windows

wherever we turn, there is the windows escaping beast...there is no escape...or should i say, there is always escape?

i don't understand, lein clean was in the original compile.bat file -- i pulled and tried building + test, and i still see an exception. am i misunderstanding?

the most recent commit is 3efab387e2842c0f5b138bdf30ededccb6b68abd

it should work with that one. see appveyor (the error was there as well, but is now fixed)

ah, running lein clean after building but before lein test? that seems to be working.

yes

ok, now i get it, thanks

i think i'm still seeing the same exceptions during the test

i am on 3efab387

this is the quote thing then?

what exceptions do you see?

i think it's the same as before, but here is a fresh copy-paste: https://pastebin.com/m0ntm5fB

sorry, for completeness: https://pastebin.com/stgtSMJk

it's the same issue:

"{:linters {:unused-binding {:level :off}, :unresolved-symbol {:level :off}, :refer-all {:level :off}, :type-mismatch {:level :off}, :unused-namespace {:exclude [.*\\\\.specs$]}}}"

on Windows

ok, thanks

at least that's what I think

lint-native! that is where the EDN config gets passed as a string

to the conch library and this uses the ProcessBuilder to make the program call

but hey, this worked in the normal JVM tests

so that's kinda weird isn't it

it is weird

oh well, no, then the string isn't handled via the processbuilder

this has nothing to do with regexp, right?

no

ok, because there seems to be something wonky about native-image and asterisk

that's a different thing

oh is that it maybe?

no idea -- but we used {0,} as a substitute for * to get native-image to succeed in working with building at some point recently, right?

i think that's a separate thing

if you put this into an EDN file:

{:linters {:unused-binding {:level :off}, :unresolved-symbol {:level :off}, :refer-all {:level :off}, :type-mismatch {:level :off}, :unused-namespace {:exclude [.*\\\\.specs$]}}}

and it seems the slashes are doubled somehow

huh

wait, the dollar sign is also weird

ah, it has to do with quotes probably

the quotes are missing

i don't seem them, yes

isn't the whole thing (from left curly to right curly) in quotes?

let me just print the config before it goes through

ok, so this works:

(edn/read-string "{:linters {:unused-binding {:level :off}, :unresolved-symbol {:level :off}, :refer-all {:level :off}, :type-mismatch {:level :off}, :deprecated-var {:exclude #:foo.foo{deprecated-fn {:namespaces [foo.bar \"bar\\\\.*\"], :defs [foo.baz/allowed \"foo\\\\.baz/ign\\\\.*\"]}}}}}")

but somehow the quotes disappear

https://bugs.java.com/bugdatabase/view_bug.do?bug_id=6511002

not sure if that's related but maybe the quotes have to be doubled for Windows?

ah this looks like it: https://bugs.openjdk.java.net/browse/JDK-8131908

could be - am reading - fwiw, there is a quoting tips section in the clj-on-windows tools.deps wiki page

Sun/Oracle refers to this document: https://docs.microsoft.com/en-us/cpp/cpp/parsing-cpp-command-line-arguments?redirectedfrom=MSDN&view=vs-2019

looking

i roll my eyes at the response at the bottom of JDK-8131908

how is anyone supposed to know what is sitting inside processbuilder on windows?

there is no mention of "windows" nor "quote" at: https://docs.oracle.com/javase/7/docs/api/java/lang/ProcessBuilder.html

nor "escape"

sorry, getting carried away 🙂

didn't lread summarize some quoting issues related to windows in some jira issue?

so this whole thing is because a string is being passed on the command line, right?

yes, I now tried something

but I didn't test it on Windows yet, so pushed it to appveyor

it worked 🙂

hurray!

all tests passing now!

will do come clean up tomorrow and then merge

awesome! > If an odd number of backslashes is followed by a double quotation mark, one backslash is placed in the argv array for every pair of backslashes, and the double quotation mark is "escaped" by the remaining backslash, causing a literal double quotation mark (") to be placed in argv. i guess you did this?

I did this:

config (if windows? (str/replace config "\"" "\\\"")
                    config)

(damn slack with their new markup thing)

i was hoping there was a way to revert -- i guess not

time for rest perhaps 🙂

oh, only a few more entries on this thread

and i can reach 100

😉

how many now?

ah nice, we broke a record!

lol, i can only easily tell that it's over 100