When writing a regular expression to match the following text.

d = """num interesting lines: 3
1
2
3
foo"""

# I only want to match the interesting lines.

m = re.match(".+?: (\d+)\n((?:.+\n){\1})", d)
print(m)
# prints: None
# Expected a match object.
print(m.groups())
# Causes Exception.
# Expected: ('3', '1\n2\n3\n')

# Works with hard coded match length.
m = re.match(".+?: (\d+)\n((?:.+\n){3})", d)
print(m.groups())
('3', '1\n2\n3\n')

Workaround it to have two regular expressions.  one to extract the desired length the second to extract the interesting lines.

I don't know of any regex implementation that lets you do that.

The RE compiler will not error out, with a back reference in there...
It treats the {\1} as a literal {\1} in the string.

In [180]: re.search("(\d) fo.{\1}", '3 foo{\1}').group(0)
Out[180]: '3 foo{\x01}'

Yes.

If it's not a valid repeat, then it's treated as a literal.

Perl does the same.

By the way, "\1" isn't a group reference; it's the same as "\x01". You should be either doubling the backslashes ("\\1") or using a raw string literal (r"\1").