Able to reproduce on Windows 7, Ubuntu 14.04 and Mac OS 10.11.5 using chrome stable M52-52.0.2743.49.

Bisect Information:
=====================
Good build: 51.0.2673.0 
Bad Build : 51.0.2675.0 

Change Log URL: https://chromium.googlesource.com/chromium/src/+log/dbdf20c9eec45a535156daf94b94db89ace868f5..7927861fcaad41e2d4d4fb713745778bde1a2b0e

From the above change log suspecting below change

Review URL: https://codereview.chromium.org/1773403002

pdr@ - Could you please check whether this is caused with respect to your change, if not please help us in assigning it to the right owner.

Thanks!

Thanks for taking the time to file this. We did change this due to https://github.com/w3c/svgwg/issues/65 and the spec has been clarified to work like you want (and not how I implemented it). My reading of the spec is that font-variant-ligatures should switch between the current behavior and the previous behavior, and the default should be the previous behavior.

Quick question: in your product, what would you want the result to be for the word "fin" where "fi" forms a ligature and therefore a single character? Assuming all characters have length 10 and the combined ligature "fi" has length 15, would you want:
length of f: 7.5
length of i: 7.5
length of n: 10

or

length of f: 15
length of i: 0
length of n: 10

or

length of f: 15
length of i: 15
length of n: 10

This expectation conflicts with what CSS WG resolved, I think CSS and SVG need to be consistent.

@kojii, I'm sorry, I've misrepresented the SVG resolution. The updated spec for getSubStringLength and getExtentOfChar is pretty clear and does not depend on font-variant-ligatures:
https://svgwg.org/svg2-draft/single-page.html#text-__svg__SVGTextContentElement__getSubStringLength

Can you double-check the spec text and see if you're happy with it compared to CSS?

Looking at the resolution in https://github.com/w3c/svgwg/issues/65, it's all about ligatures, correct? That's ok then.

This issue is different that the expectation says:
> The width of the ZWNJ should be zero
and ZWNJ forms a typographic character unit (a.k.a. grapheme cluster) with previous character. I'm not positive to get partial width of a grapheme cluster.

CSS WG has a resolution for the behavior of grapheme cluter, see:
https://github.com/w3c/csswg-drafts/commit/7dd11832ceed93cbd5ad1472d16ee24e496012f0

#5: pdr@, SVG WG and your comment #2 all talking about ligatures, while the original description is talking about grapheme clusters. Which are we really talking about?

For ligatures, CSS decided not to define the behavior atm, but in the long run, it's probably correct to get width of part of ligatures. It's just not easy and not likely to happen to us in short term.

For grapheme clusters (ZWNJ and such), I support CSS WG resolution in my comment #6.

Moving this nonessential bug to the next milestone.

For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot

Ping on this one, since it's listed as a Q2 goal for paint-dev.