Nothing should ever look at them on failure, but in some cases,
the unit tests don't check for failure, and then GCC-LTO freaks out.

Fixes part of 27772.

These confused GCC LTO, which thought they might be used
uninitialized.  I'm pretty sure that as long as 'res' indicates
success, they will always be set to something, but let's unconfuse
the compiler in any case.

This is now officially an antipattern: please let's never copy a
function declaration in two places again.  That's what headers are
for.

Part of #27772

Report what program produced the mysterious core file that we
occasionally see on Travis CI during make distcheck.  Closes ticket
28024.

This package conflicts with a preinstalled mingw openssl.

Fixes bug 27943. I think this was an appveyor change, not a Tor bug.

Closes ticket 27765

This commit also changes the appveyor file to a unified line ending
style, since the old one was pretty confused.  Use "-b" to see the
real differences.