1279467 - failing NULL byte pattern in form validation

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Description

[Rodney Rehm]

User Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.10; rv:47.0) Gecko/20100101 Firefox/47.0
Build ID: 20160604131506

Steps to reproduce:

create <input pattern="[\0]+"> and set its value to "0".

see http://jsbin.com/noledanoso/edit?html,js,output


Actual results:

the input element was set :invalid


Expected results:

the input element should be valid, as 0 (the number) is not the NULL byte.

Comment 1

[Rodney Rehm]

Of course the "Steps to reproduce" should've contained ^ to negate the character set:

  <input pattern="[^\0]+">

Note that Firefox used to get this right. Version 45 behaved like Chrome, Safari, IE - i.e. not considering the field invalid for value "0". We've seen this first break in 46.

Comment 2

[Tooru Fujisawa [:arai]]

apparently a bug in RegExp parser.
will fix this shortly.

Comment 3

[Tooru Fujisawa [:arai]]

Attachment: Fix null handling in RegExp character class with unicode flag.

Forgot to advance the cursor, and \0 is parsed as \0 and 0...

Comment 4

[Till Schneidereit [:till]]

Comment on attachment 8762086 [details] [diff] [review]
Fix null handling in RegExp character class with unicode flag.

Review of attachment 8762086 [details] [diff] [review]:
-----------------------------------------------------------------

Makes sense. r=me

Comment 5

[Tooru Fujisawa [:arai]]

thank you for reviewing :D

Comment 6

[Tooru Fujisawa [:arai]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/3e2b029c616d5cd48e7a6647181f392dd9d14ff8
Bug 1279467 - Fix null handling in RegExp character class with unicode flag. r=till

Comment 7

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/3e2b029c616d

Comment 8

[Tooru Fujisawa [:arai]]

Comment on attachment 8762086 [details] [diff] [review]
Fix null handling in RegExp character class with unicode flag.

Approval Request Comment
[Feature/regressing bug #]: bug 1135377
[User impact if declined]: Certain webpage that contains unicode RegExp in JavaScript code, or pattern attribute of INPUT element may behave wrongly
[Describe test coverage new/current, TreeHerder]: testes as automated test in m-c
[Risks and why]: Very low.  Added one missing operation that won't introduce any issue
[String/UUID change made/needed]: None

Comment 9

[Gerry Chang [:gchang]]

Comment on attachment 8762086 [details] [diff] [review]
Fix null handling in RegExp character class with unicode flag.

Review of attachment 8762086 [details] [diff] [review]:
-----------------------------------------------------------------

This patch fixes null handling in RegExp. Take it in beta & aurora.

Comment 10

[Liz Henry (:lizzard) (relman/hg->git project)]

Note for Ritu, I don't think we need to bring this into a dot release for 47. It regressed in 46 and if the problem were very widespread, we would likely have heard about it earlier or in duplicate bugs.

Comment 11

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/releases/mozilla-aurora/rev/74b32ea89a8c

Comment 12

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/releases/mozilla-beta/rev/2faf0d10a390

Comment 13

[Ritu Kothari (:ritu) (Inactive, please n-i to RyanVM, jcristau, or pascal)]

Comment on attachment 8762086 [details] [diff] [review]
Fix null handling in RegExp character class with unicode flag.

I don't believe this is a critical enough issue to be included in a dot release. Wontfix for 47.

Comment 14

[Ritu Kothari (:ritu) (Inactive, please n-i to RyanVM, jcristau, or pascal)]

Changing regression version to 46 as per comment 10.