Issue with UTF-8 caracters as parameters with python 3 about wampy

[noisyboiler]

thanks @titilambert
looks like a bug! I'll get this looked at over the coming days. You're welcome to raise a PR too if you'd like to help out?
otherwise I'll sort this out next weekend.
Thanks for bringing this to my attention.

[titilambert]

@noisyboiler I has been trying to fix it since yesterday, without success :(
I will continue to try, I hope you will be quicker than me :)

[oddjobz]

This vaguely rings a bell, can someone with the problem humour me and tell me what this comes back with?

python -c 'import sys; print(sys.getdefaultencoding())'

[titilambert]

@oddjobz

$ python -c 'import sys; print(sys.getdefaultencoding())'
ascii
$ python3 -c 'import sys; print(sys.getdefaultencoding())'
utf-8

[oddjobz]

Ok, are you mixing v2/v3 with client / server? If so, that may be the issue, setting both to utf-8 could well sort the problem. Number of ways of doing this, in the past I've change the default encoding in "sitecustomize.py", see this article on stackoverflow

[titilambert]

@oddjobz Sorry, my answer wasn't clear, I'm using python 3 ONLY. I just wanted to give the difference with python 2

[oddjobz]

Ok, but the issue "looks" like it's trying to decode something in utf-8 that's been encoded in something other than utf-8. (historically this problem is typical of swapping data between v2 and v3 instances) I don't suppose you're specifically changing the encoding anywhere in your code?

[noisyboiler]

@titilambert
can you try substituting one of wampy's existing test case calls with the text text='100éfa' to see if this is a wampy problem or not?
I'd like coverage of this regardless so it's something I'll try asap, but you're welcome to raise a PR that adds the coverage with a new test for this - else proves there is a bug!
thanks

[titilambert]

@noisyboiler I just saw, you did it here #44. Did you reproduce the bug ?

[noisyboiler]

yes! sorry, there is a bug in payload length with characters beyond standard ascii length.

[noisyboiler]

this was a combination of issues

• sys.getdefaultencoding() being ascii on our machines :(
• flakey serialisation that can return different things based on the above and the python version
• giant bug in __len__ implementation
• random decodes that made things work for ascii but were not properly understood

[noisyboiler]

fixed by #44