783607 - Write a test to check the precision of <input type=number>

Status: RESOLVED FIXED

(Core :: DOM: Core & HTML, defect)

Description

[Raphael Catolino (:raphc)]

Right now we use a double to hold the step attribute value. 
While most of the time the double's precision is enough, there are some cases where it's not enough and it's not possible to calculate the exact value of the allowed value step (http://www.whatwg.org/specs/web-apps/current-work/multipage/common-input-element-attributes.html#concept-input-step), e.g. for an <input type=number> with:
min=0.1,
step=0.2,
value=0.3, 
the element will be in the step mismatch validity step, even though it's actually in a valid state, because the double is not precise enough to handle the 0.1+0.2 addition.
The use of a double precision floating point number can lead to various edge case issues such as the one described above and, since the operations on the step/stepScaleFactor/stepBase/etc are not performance critical, we should use some arbitrary-precision type instead.

Comment 1

[Jonathan Watt [:jwatt]]

Bug 853525 fixed this issue, but it only added tests for <input type=range>. We could also do with similar tests for type=number.

Comment 2

[Mounir Lamouri (:mounir)]

Attachment: Patch

Comment 3

[Jonathan Watt [:jwatt]]

What about test_input_range_rounding.html ?

Comment 4

[Mounir Lamouri (:mounir)]

Attachment: Patch

Is that better? :)

Comment 5

[Jonathan Watt [:jwatt]]

Attachment: more realistic patch

I can't be completely sure that this will pass try since there was a huge amount of churn on the range code in the six weeks between v21 and v22 uplift. Might as well get review to make sure that this is ready as soon as possible assuming Try passes.

Comment 6

[Jonathan Watt [:jwatt]]

Comment on attachment 749810 [details] [diff] [review]
more realistic patch

Nice - totally just attached that to the wrong bug.

Comment 7

[Jonathan Watt [:jwatt]]

Comment on attachment 749803 [details] [diff] [review]
Patch

Yes. :) Thanks!

Comment 8

[Mounir Lamouri (:mounir)]

https://hg.mozilla.org/integration/mozilla-inbound/rev/64f43c57b69c

Comment 9

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/64f43c57b69c