t/op/string.t: New failures on Darwin/PPC

[jkeenan]

Somewhere between r49383 and r49432, we began to experience failures in t/op/string.t on Darwin/PPC.

$ prove -v t/op/string.t
t/op/string.t .. 
ok 1
...
ok 279 - Non canonical nan and inf
not ok 280 - Non canonical nan and inf
# Have: inf
# Want: Inf
not ok 281 - Non canonical nan and inf
# Have: inf
# Want: Inf
not ok 282 - Non canonical nan and inf
# Have: -inf
# Want: -Inf
not ok 283 - Non canonical nan and inf
# Have: -inf
# Want: -Inf
ok 284 - split - hll mapped
ok 291 - join: get_string returns a null string
1..291
Failed 4/291 subtests 
        (less 9 skipped subtests: 278 okay)

Test Summary Report
-------------------
t/op/string.t (Wstat: 0 Tests: 291 Failed: 4)
  Failed tests:  280-283

This may be due to one of two factors, or to a combination of both:

1) As in TT #1813 and TT #1814, the recent addition of configuration step auto::infnan without prior discussion or review; and/or

2) Three modifications made to the test file itself in r49415, r49417 and r49420.

Thank you very much.

kid51

[plobsing]

dup of TT #1813