On Wed, Dec 11, 2024, at 11:00 PM, Christoph Anton Mitterer via austin-group-l at The Open Group wrote: > I kinda lack any specific word which says that e.g. > file='&1' > ls does-not-exist 2>"$file" | grep 'No such file or directory' > cannot be tricked into duplicating the fd, but always has to write to > the file `&1`.
The ">&" operator is a token (see Section 2.10.2), so it has to be recognized during tokenization (see Section 2.3), which necessarily occurs before any parameter expansion. > it doesn't seem to me that this > follows strictly from the standard, as 2.7 says nothing about quoting > of word I'm not sure what this has to do with anything. > and 2.7.5/2.7.6 neither say that the & must be unquoted. Section 2.7 already says If any part of redir-op is quoted, no redirection expression is recognized. The '&' is part of the "<&" and ">&" operators and so must remain unquoted. There is no need for 2.7.5 and 2.7.6 to reiterate this. > Would people agree to that (in which case I'd open it as a ticket), or > is it felt, that it's already strictly(!) clear from the current > wording. I don't think any changes are necessary. -- vq