Fix #14533

[jgoldfar]

I tried to add a test case for this issue, but this appears to be a REPL-only error, at least to the extent that the test case

# issue #14533
function f14533_1()
    buf = IOBuffer()
    dump(buf, "TEST")
    takebuf_string(buf)
end
function f14533_2()
    buf = IOBuffer()
    xdump(buf, "TEST")
    takebuf_string(buf)
end

@test f14533_1() == "ASCIIString \"TEST\"\n"
@test f14533_2() == "ASCIIString \n  data: Array(UInt8,(4,)) UInt8[84,69,83,84]\n"

does not reproduce the issue. Interestingly (perhaps for a reason obvious to better-versed individuals) the output from xdump-ing to an IOBuffer differs from that just written to STDOUT. In fact, the error in #14533 can be avoided if you call dump(STDOUT, "TEST") which does not seem expected...

[jgoldfar]

See also #13825, I think

[tkelman]

ah right, messy to test these since the bug is only in the non-redirected versions

[tkelman]

would generally prefer more descriptive commit messages though

[jgoldfar]

Went ahead and replaced the commit with the same, plus more descriptive commit message.

[jgoldfar]

Looks like AppVeyor is having a hard time finishing the build; passed with the same content before the last change. Ref: https://ci.appveyor.com/project/StefanKarpinski/julia/build/1.0.12248

[timholy]

CC @vtjnash. Bump again if you don't get a response soon.

[tkelman]

Could still really use a deprecation for with_output_limit, given it broke Coverage.jl and probably others.