Sorry, I posted my first reply based on a too cursory reading of your question.
Code:
sed '/:$/!b;N;/:\n$/d' input.txt
Here's a brief explanation:
/:$/!b - if not a line ending with a colon, just skip to the end of the script and print.
N - this is a line ending with a colon; fetch the next line and glue them together.
/:\n$/d - if the combined two lines match this pattern, delete
else, print
The \n thing works differently in different versions of
sed, but if it doesn't work, try with a literal newline, with or without a backslash.
