Sed options beyond -e are not mentioned or called out for usability or unreliability. We found that -i in particular has inconsistent behavior, but the option is not strictly posix, so might be outside the scope of the autoconf documentation.
Gnu sed allows calling edit in place as sed -i 's/something/other/' a_file.txt MacOS (perhaps other BSD sed implementations?) requires a suffix for in-place editing sed -i.backup 's/something/other/' a_file.txt I don't know what the correct recommendation is, but personally, we are going to avoid using -i. We are definitely not the first to find this: https://stackoverflow.com/questions/69438966/i-without-argument-is-gnu-sed-posix-option-bugged-or-bsd-sed-is-not-posix-co That led me to the posix definition: https://pubs.opengroup.org/onlinepubs/9699919799/utilities/sed.html sed - stream editor - Open Group<https://pubs.opengroup.org/onlinepubs/9699919799/utilities/sed.html> APPLICATION USAGE Regular expressions match entire strings, not just individual lines, but a <newline> is matched by '\n' in a sed RE; a <newline> is not allowed by the general definition of regular expression in POSIX.1-2017. Also note that '\n' cannot be used to match a <newline> at the end of an arbitrary input line; <newline> characters appear in the pattern space as a result of the N ... pubs.opengroup.org only -e -f -n are defined. The autoconf documentation mentions -e is not consistently implemented and not to rely on its behavior. However -f and -n are not mentioned. Are those consistent? Are other commonly used flags inconsistent? The autoconf docs do not have an opinion. -i is not mentioned in Posix, and so isn't completely reliable. I would suggest a callout similar to that of grep noting what flags (if any) are safe. If a non-posix example is needed, the above would be a reasonable example. I'd be happy to help in any way I am able. Thanks! [cid:79e77e05-c6f2-475d-ae7e-becf2f43bbcc] Daniel Locks Lead Software Engineer AMA Plaza 330 N. Wabash Ave., Suite 39300 Chicago, IL 60611-5885 M: (312) 969-0217 daniel.lo...@ama-assn.org<mailto:daniel.lo...@ama-assn.org>
