Fork me on GitHub
Slack Archive

This page is not created by, affiliated with, or supported by Slack Technologies, Inc.

2022-03-16

Channels

#clojurescript
<
2022-03-16
>
Robert Brotherus09:03:28

I am a long-time user of re-frame and now wanted to try to the Karma-integration with cljs.test. Just created new project from the re-frame template with +test: 

lein new re-frame verto_gui +garden +10x +test
Then, according to the instructions in the README.md, started auto-reload and karma in another cmd: 
npm install
npm run watch
...
karma start
I see the tests being executed ok: 
Chrome Headless 99.0.4844.51 (Windows 10) verto-gui.core-test fake-test FAILED                                                       
        FAIL in   (fake-test) (verto_gui/core_test.cljs:7:9)
        "fake description"
        expected: (=
    ...
However, when I make any change to the core_tests.cljs, eg. editing "fake description" to "fake description 2", the subsequent reload produces error: 
16 03 2022 09:29:47.135:INFO [filelist]: Changed file "C:/dev/verto_gui/target/karma-test.js".
(node:5640) [log4js-node-DEP0004] DeprecationWarning: Pattern %d{DATE} is deprecated due to the confusion it causes when used. Please use %d{DATETIME} instead.
Chrome Headless 99.0.4844.51 (Windows 10) ERROR
  Uncaught SyntaxError: missing ) after argument list
  at karma-test.js:33653:588
I wonder if there is something I can do to prevent this? Or is this some bug in the framework? Since this happens with completely vanilla re-frame project from the template, it should be easily reproduced if someone is interested.

p-himik09:03:03

Are you sure the file you saved has all () paired? Because the error certainly looks like there are more ( than there are ). And I cannot reproduce that error by making a proper change to the file.

p-himik09:03:31

Although, that error is not a compilation one, so it's probably not about balancing CLJS parens...

👍 1
p-himik09:03:19

I would try opening karma-test.js to see what's at that location.

p-himik09:03:58

If you make another change, does the behavior repeat?

p-himik09:03:08

As a guess, it could've been a race condition if the JS bundle is written to disk gradually. So the compiler starts writing JS, Karma detects the change and loads the file - all while that JS is still being written.

Robert Brotherus10:03:24

With some additional tries, there seems to be indeed some randomness. On most tries I get the same error about parenthesis, occasionally another syntax error: 

Chrome Headless 99.0.4844.51 (Windows 10) ERROR
  Uncaught SyntaxError: Unexpected end of input
  at karma-test.js:47546:3

  SyntaxError: Unexpected end of input

Robert Brotherus10:03:01

And indeed in some cases works also without error. So could be race-condition that is surfacing on the specific timing characteristics of my system

p-himik10:03:31

Just in case - even though Slack embeds a preview of that message, you should still follow that link as there's a workaround posted right below the message.

Robert Brotherus10:03:20

I tried setting autoWatchBatchDelay to high value (10000 ie 10 secs) but in this case it does not seem to resolve the problem. 

16 03 2022 12:27:19.815:INFO [filelist]: Changed file "C:/dev/verto_gui/target/karma-test.js".
(node:12476) [log4js-node-DEP0004] DeprecationWarning: Pattern %d{DATE} is deprecated ...
**** 10 sec delay here ****
Chrome Headless 99.0.4844.51 (Windows 10) ERROR
  Uncaught SyntaxError: Invalid or unexpected token
  at karma-test.js:43396:1
  SyntaxError: Invalid or unexpected token
The locations and types of syntax errors still vary randomly but inspecting karma-test.js shows no visible errors at the reported location.

Robert Brotherus10:03:22

To eliminate possible sources from dangling processes and IDE (I use IntelliJ + Cursive) I rebooted and started the build with minimal environment and no IDE. With the "npm run watch" and "karma start" running I get from subsequent minimal modifications to the test-code still errors at random locations: 

16 03 2022 12:47:18.458:INFO [filelist]: Changed file "C:/dev/verto_gui/target/karma-test.js".
Chrome Headless 99.0.4844.51 (Windows 10) ERROR
  Uncaught SyntaxError: Unexpected end of input
  at karma-test.js:15043:19
  SyntaxError: Unexpected end of input

16 03 2022 12:47:38.697:INFO [filelist]: Changed file "C:/dev/verto_gui/target/karma-test.js".
Chrome Headless 99.0.4844.51 (Windows 10) ERROR
  Uncaught SyntaxError: Unexpected end of input
  at karma-test.js:10802:25
  SyntaxError: Unexpected end of input

16 03 2022 12:47:53.410:INFO [filelist]: Changed file "C:/dev/verto_gui/target/karma-test.js".
Chrome Headless 99.0.4844.51 (Windows 10) ERROR
  Uncaught SyntaxError: Unexpected end of input
  at karma-test.js:19463:1
  SyntaxError: Unexpected end of input

Robert Brotherus11:03:54

Thanks for the suggestions @U2FRKM4TW! At least we have some understanding on the nature of the issue now. I am using Windows 11, but perhaps there are similar caching issues

👍 1
Robert Brotherus11:03:10

Pity the problem does not reproduce as readily with other users as I thought, but I understand that timing-related issues can depend in a subtle way on the characteristics of ones computer, OS, etc

p-himik12:03:42

In addition, that's something that other users might not even be using at all. E.g. my UI tests are run with shadow-cljs during local development. And even if I were using Karma for CI, it would not be a problem as well because there wouldn't be any watchers, thus no chance of race conditions.

👍 1
Robert Brotherus12:03:57

Ah! The usePolling: true seems to fix the problem for me! 😀 This was hinted by the second link https://github.com/karma-runner/karma/issues/1113#issuecomment-50381150 . Although the text there suggested that Karma reverted to the polling as the default behavior, perhaps they have re-reverted to events later and therefore usePolling: true can still help (although that parameter does not seem to be even documented at http://karma-runner.github.io/6.3/config/configuration-file.html ) Thanks even more!

p-himik12:03:25

Great! Glad you got it figured out.

Simon11:03:48

Would it be possible to connect the REPL to a production CLJS app? https://twitter.com/simonelnahas/status/1504058143315308544

lilactown17:03:02

you could ship self-hosted cljs, but that dramatically increases the bundle size

p-himik17:03:36

> to e.g. emit re-frame events? If you have a very narrow scope of what you want to do in such a REPL, you can simply define and export it as a set of CLJS functions that you would then be able to call from JS.

walterl21:03:03

Say I have a server rendered Hiccup site, and would like to add two different CLJS scripts to two different pages. Is it possible to generate JS output in such a way that all the code that's common to both scripts are in one file, and the page-specific parts are in separate files?

walterl21:03:31

This would allow me to load and cache common.js on both pages, meaning faster load time for the other page.

thheller21:03:14

yes, this is possible

walterl21:03:38

w00t! 🎉

wombawomba22:03:43

I have a function that performs a slow computation and returns some data. How can I make this function run at compile-time instead of at runtime? Running it in Clojure is fine, so I suppose I should be able to wrap it in a macro?

wombawomba22:03:09

How would I actually do that though?

wombawomba22:03:44

just doing (def my-macro [] (my-fn)) and (def my-data (my-macro)) doesn't seem to work

clj.max01:03:32

@U15RYEQPJ What you did above should almost work. Have you tried (defmacro my-macro [] (my-fn))?

clj.max01:03:01

You can check with macroexpand-1 e.g. (macroexpand-1 (my-macro))

wombawomba09:03:17

yeah thanks, I managed to get it working