Given a string, is it possible to write a single POSIX-Extended regular expression to match all shortest substrings that start with a and end with b, but do not contain cb, cd or fg? I want to use such a regex with gensub, match or split functions (in gawk). For example,

1. string "0a3cbsbtacc12bbb": matching substring is "acc12b";
2. string "a4cdddbbb5": no matching substring;
3. string "1taa/b///fafgfgcb2abb": matching substrings are "a/b" and "ab".

The “mark the negative matches” approach (using characters that do not occur in the whole text) is not suitable in my situation: I am only interested in a single regular expression. So, if I use, for example, split("1taa/b///fafgfgcb2abb", arr, regex, seps), I expect seps[1] == "a/b" and seps[2] == "ab".

If it is not possible to write a suitable regex (or if any suitable regex is too inefficient performance-wise), I would be interested to know why.