Re: master 9dc306b1: Improve reporting of I/O, access errors

[Paul Eggert <eggert@cs.ucla.edu>]

On 9/18/19 12:41 AM, Michael Albinus wrote:
> Tramp trusts on file-attributes returning nil, when a
> file does not exist or is not accessible for whatever reason. This is
> true also for local files.

Could you give an example of where Tramp does this? That is, suppose 
there is a serious local-file error (I/O error, ancestor-directory 
permissions failure, etc.) when one uses Tramp to access a file - when 
would it be wrong for Tramp to signal such an error?

You suggested that there might be such an example in Bug#37202; I didn't 
quite understand that suggestion, and have followed up there.

> Furthermore, Tramp does not report anything but nil if a remote file is
> not accessible due to I/O errors. So the promise of this change cannot
> be kept for all situations.

It would be nicer if serious errors were also signaled for remote files, 
but that's not essential; such a feature can be delayed until a later 
version (if ever).

Although the jury might still be out on the 
ancestor-directory-permissions aspect of the change (certainly on 
MS-Windows, which reports EACCES for missing directories), surely Emacs 
should not simply ignore I/O errors and the like. And the permissions 
change has been helpful for testing, as it's uncovered a configuration 
bug in how Emacs starts up (Bug#37445, still being investigated). As 
that bug is possibly security-relevant, I'd like to shake it out (along 
with any similar bugs that crop up) before deciding any withdrawal of 
the ancestor-directory permissions aspect of the change.