It seems that "font-weight: lighter;" is now interpreted as "font-weight: WAY lighter".  I had to change my site to use "font-weight: 300;" to work around the fonts becoming too faint to read all of a sudden on Chrome+Android.

I'm a bit suspicious of the 'computed' weight of 200. As far as I can tell, a default build of Android will have 100, 300, 400, 500, 700, and 900 weights. The only way I think you'd end up with an actual 200 weight is if it's actually a web font. At the bottom of the 'Computed' pane in the inspector in the 'Rendered Fonts' section does it claim it a local file? In the '52' screen capture it rather looks like Roboto 100, which is what you'd actually end up with when requesting Helvetica 200.

Looking at https://www.w3.org/TR/css-fonts-3/#bolderlighter , apparently this was supposed to be 100 by the specification (I have to disagree with that table, it's a little crazy, but maybe its that way because so few fonts have different weights).

In any event, the list of changes that caused this is

https://skia.googlesource.com/skia/+/11a77c6e0634e2feb6fe4e74806db2fdd2a799ec
https://chromium.googlesource.com/chromium/src/+/932bd23be9a6b3738ec2f3593fa2fdf8326d5ab2
https://chromium.googlesource.com/chromium/src/+/f24ce736e38fa9b159d59dab44b7cc2c69a67849
https://skia.googlesource.com/skia/+/ed2edabd07086bbf60df17ca0bf52d8ba49f2273

where the third one above would be 'blamed' for this in Chromium. Essentially, 'font-weight: lighter' wasn't working at all if the initial weight was less than 600. Mobile FireFox on Android probably acts the same way Chromium used to, it may end up acting like new Chromium in the future here.

As a result, I think this is mostly 'as expected' except for the inspector claiming a computed weight of '200'.

As far as inspector is concerned, I see 200 everywhere.  "200" is what the mercurynews CSS wrote (not "lighter") -- it doesn't seem to be a webfont and you are likely right that it was resolved to Roboto 100, but perhaps that's beyond devtools' awareness.

This weekend I also noticed another repro on the subheaders of nymag.com homepage (screenshot attached).  On that page, the font-weight: 100 is explicitly specified in the CSS under ".feedlink p".

Deleted: nymag-52-palefont.png 521 KB

	nymag-52-palefont.png 521 KB View Download

Anyway, it sounds like we changed to better respect the CSS that was specified all along, but it's unfortunate that this makes Chrome paler than all other browsers.  In particular, iOS Safari has much denser fonts (see two screenshots taken on my ipod touch of the same sites).  This might just be due to Apple's font having a higher minimum weight than Roboto.

Not sure what action to take, but this probably means we can expect a steady trickle of bugs filed against Chromium on this, until all affected pages correct their CSS.

Deleted: nymag-ipodtouch.png 273 KB

	nymag-ipodtouch.png 273 KB View Download

Deleted: mercurynews-ipodtouch.png 144 KB

	mercurynews-ipodtouch.png 144 KB View Download

Anyway, this wasn't noticed in beta channel and already shipped (that's an indication the magnitude of the problem is tolerable), bungeman@'s analysis is that the behavior is correct in principle, and there probably isn't any way to make lighter weights newly available without incurring a problem like this.  So unless a further reason comes up, let's close as working-as-intended and wait for webdevs to update their CSS.

If I understand correctly, on 51 you asked for 200 but were actually getting 400; the inspector said 200 because that was the computed css value (but obviously not the resolved value of the final selected font). In 52 you asked for 200 but actually got 100 (what the spec says should be done), the inspector continues to report 200 because that is still the computed css value.

On nymag something similar happened, they were asking for 100 but were getting 400, now they're actually getting 100.

The most dramatic differences can of course be seen on demonstration pages, like http://www.quirksmode.org/css/text/fontweight.html . On Chrome 51 there are two weights displayed, on Chrome 52 with newer Android there are 6.

It would be nice if the inspector 'Rendered Fonts' section displayed more information about the resolved font. In addition to the name it would be nice to know the actual resolved styles (especially the weight). The information appears to be collected at InspectorCSSAgent::collectPlatformFontsForLayoutObject . This part is probably a dup of issue 593584 , and interested parties should star that bug. If developers could see that they were asking for 200 and getting 400 in M51 and asking for 200 and getting 100 in M52, it would be way easier for everyone to figure out what is going on.