Direct warning to io

[timholy]

CC @tkelman

[tkelman]

i feel like there might also be a doctest that's related to this error

[timholy]

Not quite sure I follow---you mean something that's triggering the error that this is designed to circumvent?

[tkelman]

No, I think one of the doctests triggers this or a very similar error, with the error output captured and some explanation nearby. I don't know whether other doctests are in a passing state on master right now so this might not be the fastest thing to find.

[timholy]

OK to merge, then?

[tkelman]

yes, but should run the doctests because I suspect #13904 could have changed some tested-for outputs.

[timholy]

Now I get what you mean. Summary on this branch:

Doctest summary
===============
 1003 tests
   59 failures in tests
    0 failures in setup code
    0 failures in cleanup code

Master prior to #13904:

 1003 tests
   59 failures in tests
    0 failures in setup code
    0 failures in cleanup code

Seems like there's no regression, anyway.

Most of those failures are due to current julia frequently adding

    in eval at ./boot.jl:264

to the end of backtraces.