I definitely don't like this, but I also don't like this from RFC2822.

It looks like tests are currently failing, as I explicitly tested for situations like this. The miri run will fail due to newer lints (it uses nightly unlike the rest of runs), so don't worry about that. With the tests updated, the change so far LGTM.

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 97.6%. Comparing base (4a74924) to head (d60fded).
Report is 43 commits behind head on main.

@@           Coverage Diff           @@
##            main    #700     +/-   ##
=======================================
- Coverage   97.8%   97.6%   -0.2%     
=======================================
  Files         81      83      +2     
  Lines       9378    9008    -370     
=======================================
- Hits        9169    8790    -379     
- Misses       209     218      +9     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

It looks like codecov is unhappy because of two lines that it counts as being in the patch, but are not actually part of the lines I touched?

Edit: nope, codecov on mobile was not actually showing me the right lines. Fixing that now.

Okay, the only failing tests is now miri, just like you said it would!

This is ready for review afaict, maybe a bit heavy on the comments — if you want to trim those a bit I don't mind! Feel free to adjust the PR before merging it (if the changes otherwise look good).

Note: I have not encountered any separators other than " " or "T" in the wild, so time could be stricter than RFC3339 but.. I'm not sure what the rationale would be! Your pick.

@jhpratt Anything I can do to help this land?

@fasterthanlime Probably the only thing would be rebasing to get miri passing. That'll save me the ~30 seconds to do it myself. Aside from that, I've been extremely busy. I do intend on reviewing this quite soon though!

Just rebased as requested! 30 seconds is 30 seconds 😎

Thanks again! Sorry about the delay.