New version of the libidn2 patch

[Florian Weimer]

I've got a slightly updated version of the libidn2 patch.  It adds tests 
using Hebrew characters, to verify that Latin numerals can be used in 
Hebrew domain names.  This is actually a feature request from users in 
Israel, and a change from IDNA2003, which rejected such names.

Unfortunately, the Latin-1 tests now fail with the current libidn2 version:

error: addrinfo comparison failure
query: xn--iebx.example:80 AF_INET/0xc2
--- expected
+++ actual
@@ -1,3 +1,3 @@
  flags: AI_CANONNAME AI_IDN AI_CANONIDN
-canonname: xn--iebx.example
+canonname: ??.example
  address: STREAM/TCP 192.0.2.120 80
error: getnameinfo test failure (code 4, flags 0x20)
   expected host:    "xn--iebx.example"
   expected service: "80"
   actual host:      "??.example"
   actual service:   "80"
error: getnameinfo test failure (code 5, flags 0x20)
   expected host:    "xn--1-qic9a.example"
   expected service: "80"
   actual host:      "??1.example"
   actual service:   "80"
error: 3 test failures

For some reason, libidn2 does not fail the conversion, but uses the 
replacement character '?' for unencodable characters.  This will have to 
be fixed in libidn2 eventually.

The question is whether we need to work around this in glibc.  I could 
synthesize an encoding failure if the output contains '?'.  But I'd 
rather not.  Users with UTF-8 locales are probably not affected by this.

If we have the libidn2 upstream fix, I will try to get it into 
distributions.  We can also add a note the glibc release notes for 2.28 
that this fix will be needed (along with a bunch of other libidn2 and 
libunistring fixes).

Thanks,
Florian
-------------- next part --------------
A non-text attachment was scrubbed...
Name: idna-shem.patch
Type: text/x-patch
Size: 992957 bytes
Desc: not available
URL: <http://sourceware.org/pipermail/libc-alpha/attachments/20180315/e0770eab/attachment.bin>