do not lie about the strings encoding

[james-lawrence]

one possible way to address #902
this is more of pointing out the issue further than an actual proposed fix.

[jeremy]

Think this is a good idea. Pragmatic choice.

[bf4]

👍 Meanwhile, in terms of a fix, I covered all encoding cases in rspec-support

• https://github.com/rspec/rspec-support/blob/81e33d4ed3/lib/rspec/support/encoded_string.rb
• https://github.com/rspec/rspec-support/blob/81e33d4ed3/spec/rspec/support/encoded_string_spec.rb

Perhaps we/I should extract parts of that

[james-lawrence]

have some further tests/fixes. this code is broken when emails contain array header fields with multiple encodings. the gsub fails and blows up wonderfully if a string is say utf-7 encoded.

changes needed + test coming up as well

[james-lawrence]

further investigation leads me to say that the multiple encodings isn't a problem unless ruby can't correctly decode once of the encoding, at which point it should blow up.

the change to the collapse method was mainly because it was generating a weird array, sometimes incorrectly parsing the encoded values. so I stripped out the regex. Anyways collapse method is likely just for our internal use atm until the underlying problem is fixed (i.e correctly blows up on strings that can't be decoded -> ::Encoding::UTF_7, ::Encoding::CP1252)

[bf4]

Extracted https://github.com/bf4/encoded_string with rspec's blessing.. It's well-tested, but I haven't done much with it on its own. Feel free to try it out

[james-lawrence]

dead pr.