Behavior: awk multi‑character field separator differences across POSIX implementations

Answer

gawk and nawk interpret a multi‑character FS as a regular expression, while mawk treats it as a literal character class (any single character from the string). The special whitespace rule FS==" " that collapses runs of spaces/tabs is honoured by gawk and nawk but not by mawk.

Likely explanation

The POSIX awk specification does not define how a FS longer than one character should be used, leaving the interpretation to each implementation. This ambiguity leads to the observed divergence.

Confirmed facts

• gawk: FS with >1 char is a regex (e.g., FS="\t+" splits on one or more tabs). FS==" " still triggers whitespace‑run splitting.
• nawk (original AT&T awk): behaves like gawk – multi‑character FS is a regex and the whitespace rule applies.
• mawk: FS is taken literally; if longer than one character it splits on any single character from that set (character class). The FS==" " whitespace optimisation is absent, so a single space acts as a literal separator and consecutive spaces produce empty fields.

Steps to verify in your restore environment

1. Create a test string with a known multi‑character delimiter, e.g., two pipes:

echo -e 'a||b||c' | gawk -F'||' '{print NF, $1, $2, $3}'

2. Repeat the same command with mawk and nawk and compare the field count (NF) and output.
3. Test whitespace handling:

echo -e 'a   b   c' | gawk -F' ' '{print NF, $1, $2, $3}'

4. Again run with mawk and nawk; mawk will report more fields because each space is a literal separator.

Missing diagnostic detail

To give a definitive recommendation, we need to know which awk implementation is invoked by your restore script (e.g., via #!/usr/bin/awk or the system default). If the script runs under mawk, you should replace multi‑character FS with an explicit regex such as FS="[|][|]" or use FS="[ \t]+" for whitespace.