[hibernate-commits] [hibernate/hibernate-validator] 13e8ed: HV-1066 Don't use IDN.toASCII on the local part of...

[GitHub]

  Branch: refs/heads/master
  Home:   https://github.com/hibernate/hibernate-validator
  Commit: 13e8edb1c8f48eba87cba418af5f217efe0d4a5f
      https://github.com/hibernate/hibernate-validator/commit/13e8edb1c8f48eba87cba418af5f217efe0d4a5f
  Author: Guillaume Smet <guillaume.smet at gmail.com>
  Date:   2016-08-30 (Tue, 30 Aug 2016)

  Changed paths:
    M engine/src/main/java/org/hibernate/validator/internal/constraintvalidators/hv/EmailValidator.java
    M engine/src/test/java/org/hibernate/validator/test/internal/constraintvalidators/hv/EmailValidatorTest.java

  Log Message:
  -----------
  HV-1066 Don't use IDN.toASCII on the local part of the email

Using IDN.toASCII on the local part of the email was a bad idea to begin
with, forcing us to introduce a splitting logic on top of IDN.toASCII.

Executing IDN.toASCII on each chunk of a split string is not equivalent
to executing it on the whole string, potentially opening the way for
other bugs.

Thus, we ended up doing the following:
- allow the \u0080-\uFFFF character range in the local part of the email
  thus allowing local parts containing UTF-8 characters;
- only use IDN.toASCII on the domain name as it should be.

This simplifies the logic a lot and should limit the number of bugs in
this area.

In passing:
- improve the feedback in email validation tests;
- be more cautious about how we generate the domain names in tests: each
  label of a domain name must be at most 63 characters long.