librustc: back::link::sanitize support escape_utf8

[yichoi]

back::link::sanitize support escape_utf8
fix #7486

[emberian]

@yichoi is this the same escaping that other tools use and accept?

[yichoi]

@cmr i refered this code https://code.google.com/p/cdf/source/browse/trunk/ebookmarket/src/main/webapp/js/escape_utf8.js?spec=svn110&r=110

[huonw]

This could probably be done by calling char::escape_default after filtering for @ .. , (or maybe escape_unicode would be better, after filtering for the ascii characters). Possibly:

 '@' => ...
 ...
 ',' => ...
 '0' .. '9' |
 'a' .. 'z' |
 'A' .. 'Z' => ...
 c => {
     result.push_char(`$`);
     result.push_str(char::escape_unicode(c).slice_from(1));
 }

which will encode things as $u1A2B and $U3C4D5E6F, depending on their length.

[yichoi]

new commit push-forced again.
@huonw Thanks! much simper than before

[huonw]

Looks good! But I'm going to nitpick again... why is this deleted? (I literally have no idea about this, so I'm sure you know more... it just seems strange.)

[emberian]

@yichoi the reason I ask is because we follow the GNU C++ name mangling scheme (I think), so all of our symbols can be put through a tool like c++filt. it would be good if we could use the same escaping scheme as them

[yichoi]

@huonw thanks again, I would pass my mistake without your kind comment

[yichoi]

@cmr do you know how c++filt handle escaping utf8 ?

[emberian]

@yichoi it seems it does not, and that the linker does not error on utf8 symbol names, that must be an android-specific problem. this is a fine fix since it doesn't do escaping.

[huonw]

(I've r+'d this for now, I'll file a bug about "name mangling and unicode" when this goes through.)

[yichoi]

rebased because of changes of char::escape_unicode