Random import error on VFS

[JamesHutchison]

brython.min.js:1 Traceback (most recent call last):
  File "#script_a26631d1b5fe46cdbd52e07652afda4c", line 2, in <module>
    from heavy_resume_brython.animations import attach_carousal
ImportError: heavy_resume_brython.animations. Did you mean: 'undefined'?

----- different attempt ------

brython.min.js:1 Traceback (most recent call last):
  File "#script_12ed66cd94ea418cbabedd4e32c191da", line 13, in <module>
    result = attach_fading_text_marquee(sig_com, **kwargs)
  File "VFS.reactpy_bridge.py", line 18, in wrapper
    return result
  File "VFS.heavy_resume_brython.animations.py", line 43, in attach_fading_text_marquee
    aio.run(fading_text_marquee(js_id,texts,duration_ms,fade_duration_ms))
  File "#script_60466087b3eb41d199be03f752bf55fc", line 13, in <module>
    result = attach_carousal(sig_com, **kwargs)
  File "VFS.reactpy_bridge.py", line 18, in wrapper
    return result
  File "VFS.heavy_resume_brython.animations.py", line 295, in start_autoscroll
    return
  File "#script_c6b964e1b4504a13b4f94b11c0dbbdf4", line 2, in <module>
    from heavy_stack_brython.scrolling import scroll_to_top
ImportError: heavy_stack_brython.scrolling. Did you mean: 'undefined'?

For reasons that are not clear to me, I cannot reliably reproduce it, suggesting it is a race condition. This was first brought to my attention when I couldn't log in on my phone all of a sudden. I tried the help link, which has a Brython based carousal in it, and recognized that it wasn't loading either.

On my computer, things ran fine, but then I noticed after disabling cache that I was getting an import error like the one above.

I rebuilt the image and it seemed to resolve the issue, except later I'm seeing the same problem with the same image.

I can quasi reproduce the issue by going into chrome. It (mostly?) seems to happen on the phone setting, or a real phone, for whatever reason.

Maybe try these steps:

• Navigate to https://heavyresume.com/login - if the google login button doesn't load, then the issue already happened
• If not, then press the "?" button and go to "The Interactive Resume"
• If the carasoul at the top doesn't load, then you have reproduced the issue. You should see the errors in the console.

The way the site uses Brython is that scripts are sent using an executor over a websocket. These scripts may include imports. It's essentially a <script> tag that imports a function then executes it.

Using Chrome 140.0.7339.128 and updating to 140.0.7339.186 did not resolve the issue

I believe this is 3.13.2

[JamesHutchison]

I'm going to take a look at this the first chance I get tomorrow (it's 8:44 pm for me) if someone doesn't beat me to it.

[JamesHutchison]

Something's not right with the imports. We can see that some modules have __initialized__, some have __initializing__, and some have __path__ and some do not. The contents seem correct when imported but the meta values are not.

The underlying error occurs because __path__ is missing, not because it's not imported.

                try{
                    __path__ = $B.$getattr($B.imported[_mod_name], "__path__")
                }catch(e){
                    // If this is the last but one part, and the last part is
                    // an attribute of module, and this attribute is a module,
                    // return it. This is the case for os.path for instance
                    if(i == len - 1 &&
                            $B.imported[_mod_name][parsed_name[len]] &&
                            $B.imported[_mod_name][parsed_name[len]].__class__ ===
                                $B.module){
                        return $B.imported[_mod_name][parsed_name[len]]
                    }
                    if(has_from){ // "from a import b" : ImportError
                        import_error(mod_name)

[JamesHutchison]

I'm guessing this is the source of the bug as we're overwriting the parent object if it hasn't already gone through this logic. An assignment to __path__ is notably missing.

VFSLoader.exec_module = function(self, modobj){

    ...

    }else if($B.precompiled.hasOwnProperty(modobj.__name__)){
        if($B.get_option('debug') > 1){
            console.info("load", modobj.__name__, "from precompiled")
        }
        var parts = modobj.__name__.split(".")
        for(var i = 0; i < parts.length; i++){
            var parent = parts.slice(0, i + 1).join(".")
            if($B.imported.hasOwnProperty(parent) &&
                    $B.imported[parent].__initialized__){
                continue
            }
            // Initialise $B.imported[parent]
            var mod_js = $B.precompiled[parent],
                is_package = modobj.$is_package
            if(mod_js === undefined){
                // Might be the case if the code in package __init__.py
                // imports a submodule : the parent of the submodule is not
                // yet in precompiled
                continue
            }
            if(Array.isArray(mod_js)){
                mod_js = mod_js[0]
            }
            var mod = $B.imported[parent] = Module.$factory(parent,
                undefined, is_package)
            mod.__initialized__ = true

I'll need to dig into the __initializing__ versus __initialized__ logic as I don't think modules should have different magic variables depending on what logic path invoked the import

[lefranco]

Traceback (most recent call last):
File "#main", line 1, in
import index
File "VFS.index.py", line 5, in
from json import loads, dumps
File "VFS.json/init.py", line 120, in
from.encoder import JSONEncoder
File "VFS.json.encoder.py", line 3, in
import re
File "VFS.re.py", line 1, in
import enum
JavascriptError: TypeError: SEQ is not iterable

code is
from json import loads, dumps

Seems it happens only on firefox/Android (gecko)