md5(1): fix -c flag to work with input on stdin

[allanjude]

Summary

Previously, the -p and -c flags were ignored when reading from stdin
Additionally, -s and -c can be used together now.

PR: 247295

Test Plan

$ cat file | sha256 -c f2ca1bb6c7e907d06dafe4687e579fce76b37e4e93b7605022da52e6ccc26fd2; echo $?
f2ca1bb6c7e907d06dafe4687e579fce76b37e4e93b7605022da52e6ccc26fd2
0
$ cat file | sha256 -c dead1bb6c7e907d06dafe4687e579fce76b37e4e93b7605022da52e6ccc2beef; echo $?
f2ca1bb6c7e907d06dafe4687e579fce76b37e4e93b7605022da52e6ccc26fd2 [ Failed ]
2
$ sha256 -s foobar -c foobar; echo $?
SHA256 (foobar) = c3ab8ff13720e8ad9047dd39466b3c8974e592c2fa383d4a3960714caef0c4f2 [ Failed ]
2

[kevans]

sbin/md5/md5.c

526 ↗

(On Diff #73341)

If I'm reading this right, this is actually passing the address of buf (stack-allocated) back to the caller. I suspect we really want buf to go away and just pass in NULL, then make sure to free() the buffer in the callers.

[allanjude]

sbin/md5/md5.c

526 ↗

(On Diff #73341)

Yeah, we should use the buf[] that already exists in the calling context atleast

[allanjude]

Address review feedback

[kevans]

I think I'm happy with this revision. :-)