Runtime error (IronPython) about pysnooper

[cool-RR]

@runxel could you fetch the path that includes the illegal path? Like put a print statement just before the line that fails?

…

On Sun, Apr 28, 2019, 23:29 Lucas Becker ***@***.***> wrote: True, I will just clone it :) Thank you very much already for looking into this crude edge cases! — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#75 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAAN3SU5RR3CKVYEYGL5COTPSYCL5ANCNFSM4HI66E5Q> .

[cool-RR]

Please specify iron python version

…

On Sun, Apr 28, 2019, 20:07 Lucas Becker ***@***.***> wrote: I get the following runtime error with the demo code provided in the readme: Runtime error (MissingMemberException): 'NoneType' object has no attribute 'get' Traceback: line 72, in decorate, "pysnooper.py" line 59, in get_source_from_frame, "pysnooper\tracer.py" line 209, in trace, "pysnooper\tracer.py" line 2, in number_to_bits, "<decorator-gen-10>" line 20, in script (line 20 calls the "number_to_bits" function) — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#75>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAAN3SXTIVTZYZBF5NVI75LPSXKWTANCNFSM4HI66E5Q> .

[runxel]

@cool-RR

2.7.8 (IronPython 2.7.8 (2.7.8.0) on .NET 4.0.30319.42000 (64-bit))

(And sadly switching versions is not an option, since this is tied to the program I use (Grasshopper))

[cool-RR]

Which version of PySnooper?

[runxel]

Downloaded from Github 4 days ago.

[cool-RR]

Give me the version number, use `pip freeze`

…

On Sun, Apr 28, 2019, 22:10 Lucas Becker ***@***.***> wrote: Downloaded from Github 4 days ago. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#75 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAAN3SWHQKBIL52U2EQP7GDPSXZDLANCNFSM4HI66E5Q> .

[runxel]

Ah, I didn't use pip. Basically because you can't use pip for that.
When using Grasshopper (part of the NURBS modeling software Rhino) you have to manually download and reference any third-party modules one would like to use.

[cool-RR]

I believe I fixed it. Try the new release and please report back.

@bittner It won't be easy to set up Travis to test on IronPython, right?

[runxel]

Interestingly it now says illegal char in path.... 🤔

Runtime error (ArgumentException): Illegales Zeichen im Pfad.
Traceback:
  line 75, in inner, "pysnooper.py"
  line 110, in get_source_from_frame, "tracer.py"
  line 229, in trace, "tracer.py"
  line 2, in number_to_bits, "<decorator-gen-2>"
  line 19, in script

[alexmojaki]

@cool-RR if you're going to chase this, it might save you time not to release a version at each iteration, since @runxel is installing from GitHub anyway. Maybe even make a separate branch.

[runxel]

True, I will just clone it :)

Thank you very much already for looking into this crude edge cases!

[bittner]

@bittner It won't be easy to set up Travis to test on IronPython, right?

IronPython is Windows, and Windows environments are not provided by Travis. We can do AppVeyor, though. Just like with PythonTurtle.

[cool-RR]

@bittner there's mono for Linux though, isn't there?

…

On Mon, Apr 29, 2019, 03:51 Peter Bittner ***@***.***> wrote: @bittner <https://github.com/bittner> It won't be easy to set up Travis to test on IronPython, right? IronPython is Windows, and Windows environments are not provided by Travis. We can do AppVeyor, though. Just like with PythonTurtle. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#75 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAAN3SSZUIGSD7DRSHMSRQ3PSZBBFANCNFSM4HI66E5Q> .

[bittner]

Like put a print statement just before the line that fails?

LOL 🤣

[cool-RR]

Haha yeah, the irony is not lost on me :) And also: https://mail.python.org/pipermail/python-ideas/2019-April/056443.html If anyone's interested, it'll be cool to write a PEP for this.

…

On Mon, Apr 29, 2019 at 9:46 AM Peter Bittner ***@***.***> wrote: Like put a print statement just before the line that fails? LOL 🤣 — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#75 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAAN3SX4G7OGQYKZ64MFVI3PS2KVBANCNFSM4HI66E5Q> .

[runxel]

So, what baffles me is that I am completely lost what path it's talking about.
Also looking into your code gives me no hints to what to print.
(Or can I somehow insert something like print into your code, and get that output?)

@runxel could you fetch the path that includes the illegal path? Like put a print statement just before the line that fails?

[cool-RR]

@runxel See the line with open(file_name, 'rb') as fp: ? The file_name argument probably has an illegal character or something. Add a line just before that line to print file_name, or write it to a file.

[runxel]

Ach, totally forgot I could write to a file – my bad!
So using file.write(file_name) I obtained this content:
<string><string>
This is literally the whole text file. 😨

[cool-RR]

I'm trying to reproduce this, but I'm not getting the ArgumentException that you're getting. I didn't even know it existed. Can you catch it as e and then print more information about it, especially e.mro()?

[runxel]

Now that was a fun story:
To actually catch the ArgumentException I had to say

# line 109 ff. in tracer.py
try:
    # foo
except Exception as e:
    # pass

And by that alone the error was gone!
Which actually makes pretty much sense. You have 2 if source is None there. We are in the first if and in it there we get into the else condition – We can never reach the second if source is None.
It then tries to write to a file with a path which is nonexisting and the ArgumentException gets not handled, only OSError and IOError.
Maybe this section should be revised.

Since we can never save the actual Python script in Grasshopper but stuff it into a component (Grasshopper is a node-based visual scripting environment) there will be never a filename.

Of course the except Exception is not a smart thing to do in the real world. The culprit is already in file_name = frame.f_code.co_filename where it is assumed, that this will always give a valid file name. As mentioned above this is not always the case, since the user might run scripts in arbitrary environments.
I'm not sure if we can detect such cases correctly. The other fix would be checking file_name against <string> as content. (I had a mistake there above because I wrongly used a instead of w as the file open argument. So frame.f_code.co_filename interestingly gives this non-empty empty content.)

[cool-RR]

I think you're making a couple of mistakes there. Yes, I know we have to catch more than OSError and IOError, the question is what is that error, because ArgumentException is not a Python error.

Please try adding ValueError to the except clause at line 112 and tell me whether that solves your problem.

[runxel]

Right, there is no ArgumentException; I guess it comes from C# since Rhino/Grasshopper boils down to this in the end.
Yes, ValueError will do the job!

[cool-RR]

Great, I released a new version, please try it and update here.

…

On Tue, Apr 30, 2019 at 8:12 PM Lucas Becker ***@***.***> wrote: Right, there is no ArgumentException; I guess it comes from C# since Rhino/Grasshopper boils down to this in the end. Yes, ValueError will do the job! — You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub <#75 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAAN3SWEQREJ47ZW2XM4M5LPTB4YRANCNFSM4HI66E5Q> .

[runxel]

Works gracefully!
Thank you so much :)

[cool-RR]

Awesome, this was a fun bug :) There should also be an open bug on IronPython for raising `ValueError`, but I'm gonna let the universe handle that one :D

…

On Tue, Apr 30, 2019 at 8:36 PM Lucas Becker ***@***.***> wrote: Closed #75 <#75>. — You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub <#75 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAAN3SUZJKVS7YNLSW4JUF3PTB7TLANCNFSM4HI66E5Q> .