I'm wondering if the messages could use a slight rewording, to give a
clue to the user about the reason *why* the expression has changed
signedness. The old message "signed and unsigned type in conditional
expression" gave the clue (but failed to underline the subexpression
changing sign, and tell what the old/new types were).
A horribly verbose way to put it would be something like:
"operand of conditional expression with mixed signedness changes
signedness from %qT to %qT due to promotion to unsigned to match
unsignedness of other operand" (ugh)
(assuming I'm understanding the logic correctly)
or something like:
"operand of conditional expression changes signedness from %qT to %qT
due to unsignedness of other operand"
or somesuch (am not 100% happy with that either).
If I can make an observation I'd say (since you're not happy with
the wordiness) that mentioning signedness in addition to the types
of the operands is superfluous: it should be sufficiently clear
from the "from 'T' to 'unsigned T'" part.
That being said, what might be helpful is mentioning which operand's
type changes: the second or third. That could be viewed as redundant
as well thanks to the underlining but only as long as the underlining
is correct and as long as it's not disabled by some option.
With that I'd suggest to consider the simple:
second operand of conditional expression changes type from %qT to %qT
Martin
