Description: In order to make the output of `complete -p' reusable as shell commands, the function name after `-F' and the command names also needs to be properly quoted.
I understand that both the function names and the command names don't usually need to be quoted, but in principle it is possible to create such completion settings. This becomes a problem when the script needs to handle user-supplied arbitrary completion settings. The related issue is https://github.com/akinomyoga/ble.sh/issues/102 I need to evaluate/parse the result of `complete -p' in my script, I'm currently using a complicated heuristic way to analyze `complete -p', but really want an efficient and reliable way. i.e. something like « eval "words=($(complete -p CMD))" ». Repeat-By: For example, the output of `complete -p' in the following cases cannot be reliably evaluated. $ complete -c 'test1; echo yes' $ complete -F 'A{B,C}' test1 $ complete -p complete -c test1; echo yes complete -F A{B,C} test1 If we perform « saved=$(complete -p); ...; eval "$saved" », the above types of completion settings will cause unexpected results. Fix: I attach a patch (r0028-quote-complete-p-A.patch). In this patch, the command names are quoted only when necessary. the function names after `-F' are always quoted. If we want to quote the function name only when necessary, please instead use the second patch `r0028-quote-complete-p-B.patch'. -- Koichi
r0028-quote-complete-p-A.patch
Description: Binary data
r0028-quote-complete-p-B.patch
Description: Binary data
