Initial support of Unicode filenames

[Martin Geisler]

Antoine Pitrou <solipsis at pitrou.net> writes:

> On Thu, 03 Nov 2011 10:31:28 +0100
> Martin Geisler <mg at aragost.com> wrote:
>> 
>> That's the point: some people think it's better to checkout files
>> with broken filenames instead of refusing to checkout.
>
> The filenames weren't "broken" in the first place when they were
> checked in, it's just that (IIUC) hg prefers to faithfully copy their
> bytes representation rather than transcode them to a locale-agnostic
> representation.

Yes, that's right.

> The filenames may be unrepresentable in the reader's locale, though.
> There may be solutions to that. Not necessarily pretty ones - they
> probably involve a funky escaping algorithm -, but the current
> situation isn't pretty either: in both cases, the user gets a
> misrepresented filename (aka mojibake).

Correct.

> What seems to be the design point is that filenames stay the same in
> their binary representation (so that Makefiles aren't broken, I
> suppose).

Yes, Makefiles are often used as an argument here. Ant files provide a
counter-argument.

> Of course, that prevents any proper unicode support if computers with
> different encodings are involved. (and "unicode" really means "the
> user can read the checked out filenames")

I agree.

>> Today, a Windows user can commit a file named "Sweet crêpe
>> recipe.txt" and I can checkout the file on my Linux machine. I won't
>> get a "ê" in my filename, but I'll get a file I can modify and commit
>> changes to anyway.
>
> But proper unicode support /would/ get you a "ê" in the latin1
> filename. (you may have to use the wide APIs under Windows, if "ê"
> isn't in your local code page)

Yes, that's the goal: give users readable filenames as long as their
locale supports it. It will breaks some (very few?) Makefiles, but it
will make users who regularly use non-ASCII characters in their
filenames very happy.

-- 
Martin Geisler

aragost Trifork
Professional Mercurial support
http://mercurial.aragost.com/kick-start/