Diffenator and Diff Browsers new main results

[vv-monsalve]

Big Shoulders Text Regular

Browser Diff / Chrome 69

Comparative images

[vv-monsalve]

Big Shoulders Text Bold

Browser Diff / Chrome 69

Comparative images

[xotypeco]

the changed marks and kerning are expected; the last set of missing marks are a mistake. it looks like I forgot to fix some anchors. will fix. thanks, Viviana.

[vv-monsalve]

Big Shoulders Display Regular

Browser Diff / Chrome 69

Comparative images

[vv-monsalve]

the changed marks and kerning are expected; the last set of missing marks are a mistake. it looks like I forgot to fix some anchors. will fix. thanks, Viviana.

Yes, don't worry. It's possible some of them are expected, the images are helpful to confirm everything.

[xotypeco]

question: does this mean that the kerning pairs are being shown as completely missing? I'm going through them and in most cases (LT in particular) it's values I’ve opened, but they're still tightened.

edit: I think these kerning results are somehow incorrect—the kerns it's reporting as missing are actually present in the source files & look different from the browser results. in particular, LT is very different from source file, as are all of the kerns with uppercase T following it.

but I don't know why they would be reported wrong, which is why I'm mentioning. thought you might have some insight.

[vv-monsalve]

Sometimes it is due to exceptions on some Kerning class. So the leader letter has the right value, but some related that should have the same value is appearing as an exception and has a different one or none at all.

Did you check the source or directly the interpolated instance? Sometimes the source doesn't have any problem, but the resulting font yes. It could lead to some hints on the matter.

In any case, for the variable, it would be needed that all the masters have the same amount of kerning pairs, so if it is a kerning pair in one master but none in the other, that last should be at least '0'.

[xotypeco]

Did you check the source or directly the interpolated instance?

I checked source, but I stopped before I got through everything, to ask you. is it possible for you (or me, for that matter) to generate a text file with all those kerning pairs showing exceptions?

edit: all of the masters in LT are kerned, but marked as missing in your output gifs, so I'm hoping that points to something clearer

[xotypeco]

a little more information I just found, which might help: I was experimenting with a sample string from the first line of kerns reported as missing ( LT ĿT ĽT ḶT ŁT ḺT Tṥ Tṩ Tṡ T… T· T: ) and those kerns do actually exist & work as expected in the generated fonts from your repo fork (I tested in Adobe Illustrator.) hopefully that points to how this result is occurring.

[vv-monsalve]

I checked source, but I stopped before I got through everything, to ask you. is it possible for you (or me, for that matter) to generate a text file with all those kerning pairs showing exceptions?

Diffenator automatically generates those. Nevertheless, you can find here a rich source of strings for kerning

[xotypeco]

ok. I’m not really sure how to address this. those diffenator results appear to be literally reporting incorrect information, per my testing here.

[vv-monsalve]

a little more information I just found, which might help: I was experimenting with a sample string from the first line of kerns reported as missing ( LT ĿT ĽT ḶT ŁT ḺT Tṥ Tṩ Tṡ T… T· T: ) and those kerns do actually exist & work as expected in the generated fonts from your repo fork (I tested in Adobe Illustrator.) hopefully that points to how this result is occurring.

Ok, that might help. However, maybe a double-check on how the fonts behave on browsers could be sensible. You could test that in the same Html to test the Vietnamese. Also, Cyreal Font Testing Page displays some kerning examples.

[xotypeco]

everything works as it should at the Cyreal page (see screenshots with some carets to show where the kern happens).

here's the Vietnamese page. it's fine. no clipping that I could find.

[vv-monsalve]

Checking it in Cyreal the paragraphs seem to be ok with no-reflow happening on the text as it in Diffentor previews.

---

The Vietnamese preview on Windows browsers seems to be clipping free. Now, you might want to check letters with horn (like uhorn and horn) are touching the adjacent glyphs, which may lead to legibility issues. Also, to increase space between these diacritics, it looks like they are overlapping with each other or the base letter (e.g., ắ).

[vv-monsalve]

Beyond those Vietnamese improvements, I think it would be ok for the update, however, I would like to know @m4rc1e thoughts.

[xotypeco]

I'm working on the Vietnamese now. (I made a fork of your fork to catch the changes you made, just as an FYI.) I'm changing the hook by moving it up & making it ia little bigger in lighter weights; the size is a mistake. it looks like a little spot at small sizes.

[xotypeco]

ok, all this is fixed up and all changes merged to master. Vietnamese also updated at proofs here.

[vv-monsalve]

(I made a fork of your fork to catch the changes you made, just as an FYI.)

Thanks for taking my changes. Just to be sure, did you take the most recent work done on QA-BigShoulders branch?

[xotypeco]

Just to be sure, did you take the most recent work made on QA-BigShoulders branch?

yes!

also, I forgot to say: I'm working the Vietnamese changes we just talked about into Stencil as well.

[vv-monsalve]

Ok, great! I wanted to be sure before pulling your changes.

Here are comparative gifs for Vietnamese. Please check everything is as expected.

Chrome 83 - Win 10

Firefox 77 - Win 10

Edge 83 - Win 10

Opera 68 - Win 10

[xotypeco]

yes, all as expected. thanks!

[xotypeco]

a little more info on why the widths with horns are too narrow:

in previous versions, masters of Thin and Black had some offsets to right siebearing set to account for the horn, but the new Regular master made those values collapse back to the base character's right sidebearing, even though the offsets were present.

it looks like the offsets simply weren't carried over into new interpolations, and weren't reflected in regular at all. I needed to re-set them.

(just in case you run into it again in future work.)

[xotypeco]

here's Stencil with updated Vietnamese, for QA. I also pushed new source including these changes. it'll happen once more when I get to Inline, so it might be a good idea, for the sake of streamlining, to hold off on branching to QA again, just so you're not duplicating effort.

can you detail the changes you'd made to regular Big Shoulders so I can duplicate that in here? or just point to your original comment? can't find it in search. thanks!

[vv-monsalve]

In that case, I think the main changes in the source files would be:

• To ensure the "axes coordinates" value for the regular master matches the value of the instance (to avoid conflicts both with interpolation and Vfs later). Currently, the master is 300, which creates a conflict with Light instance (which is also 300) and with the regular instance (which is 400). It would be necessary to run "instance as master" and take that new master as the new Regular one.

• Checking the kerning pairs and ensure all the masters have the same quantity of classes and exceptions, even if for some cases the value is just 0. It is particularly needed to build the Vfs.

• Finally, to change the instance names to Text and Display to follow wha Dave said on releasing steps.

Having that on the files I'll continue from there with the other fixes and create the needed build scripts.

[xotypeco]

To ensure the "axes coordinates" value for the regular master matches the value of the instance (to avoid conflicts both with interpolation and Vfs later). Currently, the master is 300, which creates a conflict with Light instance (which is also 300) and with the regular instance (which is 400).

actually this is a case of me saying the wrong thing.

I forgot that 300 is actually Light instead of Regular.

for a long while the weights were 100 (thin), 200 (light), 300 (reg), but at that point Marc was trying to work out (for all of Google Fonts) what should happen to UltraLight and Thin fonts. either he or yanone re-sequenced them to 100 (thin), 300 (light), 400 (reg), and I forgot that had happened.

(what you're suggesting needs to be done is correct. the explanation is to shed light on how the mistake happened so that I seem less like an idiot. lol.)

[vv-monsalve]

Yes I understand, don't worry. I even looked for some Issues regarding this to share them with you and a couple of them are related to Big Shoulders actually, lol

As reference is better now tho inspect the Specifications Doc that Marc is building, which gathers the most up do date standards, at least to contribute to Google Fonts :)

[xotypeco]

and a couple of them are related to Big Shoulders actually, lol

honestly, I'm hardly surprised by this

[vv-monsalve]

yes, all as expected. thanks!

Great, so can we go ahead with this variant? Is the source file on your master the latest, ready to resume QA?

[xotypeco]

I need to double check all those kerns one more time to make sure everything interpolates. i had forgotten that from your first batch of findings. will let know know when done; I think I can get them solidified tomorrow.

[vv-monsalve]

Ok, let me know

[xotypeco]

ok, Big Shoulders source pushed to master with all those edits. one edit further: I set the origin to text light rather than to add another master.

Big Shoulders Stencil is also pushed, but that one's not final yet. it needs kerning to be equal in all masters. I'm doing that tomorrow & will let you know when it's ready.

[vv-monsalve]

ok, Big Shoulders source pushed to master with all those edits. one edit further: I set the origin to text light rather than to add another master.

Ok, great! I'll pull it now

Big Shoulders Stencil is also pushed, but that one's not final yet. it needs kerning to be equal in all masters. I'm doing that tomorrow & will let you know when it's ready.

Good. what would it be for Big Shoulders Inline?

[xotypeco]

Good. what would it be for Big Shoulders Inline?

probably Tuesday at latest.

for that one, I need to fix the missing diacritics you'd found Big Shoulders, doublecheck Vietnamese hook mark (I think that one's fine), Uhorn and Ohorn sidebearings, and import metrics from Big Shoulders.

[xotypeco]

Stencil updated for you. on to Inline next.

[xotypeco]

surprise, Inline is also pushed! it turned out to be substantially easier than Big Shoulders or Stencil to update metrics in the Vietnamese because the base characters with horns weren't made of base-and-accent pairs. the horn is drawn connected to the base in that character.

I think you have everything you need; lmk if you run across anything amiss and I’ll correct it asap. Vietnamese proof for Inline coming shortly.

edit: hold on Inline. gotta re-check kerns.

[xotypeco]

ok, got kerning fixed. the first import gave me some really weird results.

Vietnamese testing is here.

[vv-monsalve]

Latest Browsers previews

OS X Sierra Safari 11

Windows 10 Chrome 69

Windows 10 Firefox 62