POSIX behavior of sed's ^ and $ regex anchors with multiline pattern spaces

Question

What is the POSIX behavior of the sed ^ and $ anchors when used at the beginning and end of a regex if the pattern space contains embedded newlines, like after you use the N command?

Should they still match the beginning and end of the pattern space or should they match the beginning and end of any line now that it contains multiple lines?

GNU sed's ^ and $ still match only the beginning and end of the entire pattern space.

Answer 1

^ and $ in BRE always match at the start and end of line, as describe here.

Any sed which use BRE will do the same way. In case of using N command, sed saw multi lines as one long line in pattern space, with each real line separated by embedded newline \n (literal \ and n).

For confirmation:

printf '1\n2\n' | sed '1N;/2$/d'

output nothing. And:

printf '1\n2\n' | sed '1N;/1$/d'

gave you:

1
2

Or using look command for more verbose:

$ printf '1\n2\n' | sed 1N\;l
1\n2$
1
2

$ printf '1\n2\n' | sed 1N\;l | od -t a
0000000   1   \   n   2   $  nl   1  nl   2  nl
0000012