2023年2月4日(土) 20:43 Oğuz İsmail Uysal <oguzismailuy...@gmail.com>: > On 2/4/23 12:23 PM, Koichi Murase wrote: > > Changing the behavior related to the function names wouldn't make the > > behavior of the shell entirely unspecified > I see, but that's not what you're suggesting. You're suggesting that how > command search and execution works be changed to allow functions that > contain slashes in their names to prevent command search, and shadow > filesystem entries including those of standard utilities.
I think the shells that do not support defining such a function are not affected, so do not need to change the search & execution model. Even the shells that allow defining such functions do not need to really change the behavior because it is just explicitly stated as `unspecified'. > The fact that bash and zsh already allow it doesn't mean it's a good > idea. I think the last sentence of my second last reply to you [1] would be the reply to this sentence. I'm not trying to argue that it's a good idea based on the existing implementations, Bash and Zsh. Even if we would forget about the existing implementations, I'm not sure if it would absolutely be a good idea to disallow calling functions with slashes in their names. [1] https://lists.gnu.org/archive/html/bug-bash/2023-02/msg00024.html Anyway, the reason that I raised Bash and Zsh is for backward compatibility. I haven't argued whether the behavior is good/reasonable or not until you mentioned it, but I have been discussing compatibility from the beginning. I first mentioned Zsh for the compatibility with different shells that the POSIX seems to care about. > And it's easier to make the overall behavior unspecified than > examine each shell and document this new behavior; I thought we can just easily append a sentence to [2.9.1 / Command Search and Execution / rule 2] like « When the implementation extends the character set allowed in the function name to include a <slash> as described in [2.9.5 Function Definition Command], if the command name matches an already-defined function name, it is unspecified whether rule 1c is applied or rule 2 is applied. » It basically loosens the requirement by the standard so that the behaviors of Bash and Zsh are included, so I don't think we need to examine each shell for the new requirement. But I might miss something non-trivial. -- Koichi
