bug#22145: 25.0.50; tramp and auto-revert tests hang on OS X

[Michael Albinus]

Random832 <random832@fastmail.com> writes:

Hi,

> If I were to guess that the Chinese name works fine and the Greek and
> Cyrillic ones do not (don't know enough about the script to make an
> immediate guess about the Arabic), would I be right? Or can the test
> failure not give you that information?

The Arabic name let's the test freeze forever; it is sklipped for OSX
already. From the other names, my first candidate for producing the
problem is the Chinese name. I've suppressed it as well earlier today,
let's see the feedback from Anders or somebody else.

> Have you been following the discussions last month about the issues with
> file name completion for local OSX filenames?

Yes.

> The issue is most likely that filenames on OSX are  (in most
> circumstances) normalized to turn diacritics into combining characters.
> This is undone, on OSX machines, by using the utf-8-hfs (formerly
> utf-8-nfd, which was incomplete and may or may not have worked for your
> purpose, but the patch for last month's bug replaces it with an alias
> for utf-8-hfs) encoding.

But IIUC correctly, this touches only local file names and how they are
handled in Emacs. Tramp works on the remote side.

Anyway, this is exactly something I'd like to debug.

>> Likely, I could test it without running Emacs on an OSX machine. If
>> somebody could give me temporarily a very restricted account on an OSX
>> machine, I could use "/ssh:restrictedaccount@osx-machine:/tmp" as test
>> target. If the problem is reproducibly with this, I could debug Tramp,
>> running on my local machine.
>
> I'll see what I can do this weekend.

That would be great!

Best regards, Michael.