Bad character escapements for some bold fonts

[hgsfghs]

BFont.StringWidth () and BFont.GetEscapements () seem to return overly wide sizes for bold fonts. It can be reproduced by the following code:

    BView vw; /* Get a view somehow */
    const char *str = "Haiku test string";
    float fl[strlen (str)];
    edge_info ei[strlen (str)];
    ft.GetEscapements (str, strlen (str), (float *) &fl);
    ft.GetEdges (str, strlen (str), (edge_info *) &ei);
    vw->LockLooper ();
    vw->PushState ();
    vw->PushState ();
    vw->SetHighColor (128, 0, 0, 255);
    vw->SetPenSize (2.0);
    BRect r (0, 0, ft.StringWidth (str) - 1, 50);
    vw->StrokeRect (r);
    vw->PopState ();
    vw->MovePenTo (0, 50);
    vw->SetFont (&ft);
    vw->DrawString (str);
    vw->PopState ();
    vw->UnlockLooper ();

The rectangle drawn will be far wider than the text drawn, which is not the case with a plain font. Similar issues, albeit on a character-by-character basis can be seen in the escapements returned by BFont.GetEscapements ().

I have attached two screenshots, depicting the difference between the result of calling GetStringWidth () on a regular font and that of calling the same function on a bold font.

[hgsfghs]

StringWidth of string with bold font

[hgsfghs]

StringWidth of string with plain font

[waddlesplash]

madmax, you seem to be the burgeoning fonts rendering person, perhaps you might have some time to look into this?

[madmax]

Reproducible with Noto Sans Bold, but not with Noto Sans Display Bold (or any other of the very few I've tested with).

[hgsfghs]

I also noticed that BFont.GetEdges () and BFont.GetEscapements () doesn't seem to respect shear. Could this be related somehow?

[madmax]

The problem is not with metrics. The painter is using SemiBold style instead of Bold for that font. Will try to find out why later in the evening.

[madmax]

Patch: ​https://review.haiku-os.org/c/haiku/+/4416

[hgsfghs]

Thanks for the help and timely fix :)

[hgsfghs]

It doesn't seem like the patch has been installed yet. Is there any way I can help with this? Right now, it's preventing some important features such as reading e-mail with RMAIL from working correctly in Emacs.

Thanks.

[pulkomandy]

Sorry, it seems this one got lost in the code review process, I have just merged it now.

[hgsfghs]

Thanks! One last question: will this patch end up in R1/Beta3 systems updated through SoftwareUpdater, or will it only show up in nightlies?

[pulkomandy]

I have only merged it for nightlies. We can backport it to the beta3 branch as well if needed.

[hgsfghs]

It would be nice to have Emacs work properly in R1/Beta3 (since that's the system I use for testing), but if you think it's too risky to backport then that's fine as well.

Thanks :)

[pulkomandy]

I'd say let's wait for a week or two and if no one complains about regressions in nightlies we can backport it. Remind me then because I will probably forget to do it.

[hgsfghs]

Okay, thanks!

[hgsfghs]

It's been two weeks, so if there haven't been any regressions I think it would be nice for someone to backport it to R1/Beta3. WDYT?

Thanks.

[korli]

Backport change: ​https://review.haiku-os.org/c/haiku/+/4896

[madmax]

Replying to hgsfghs:

It would be nice to have Emacs work properly in R1/Beta3 (since that's the system I use for testing)

To be clear, does that mean we have not checked the patch solves your real problem in nightlies? It does fix a problem with font flags that leads to the issue in the description, but I only ran the test you provided and a few similar ones. Is that extracted from the code used by emacs? If that's not the case, can you provide instructions (detailed ones including setup, I have no emacs experience) to reproduce it in emacs?

[hgsfghs]

No, I just meant to say that it would be nice to have it backported to the beta branch, since I use that for testing. I installed one of the nightlies to test, and it does fix the problem with Emacs.

Thanks.

[korli]

This change is backported to the beta3 branch in Git.

[hgsfghs]

Thanks.

[korli]

can you confirm it works as expected for you on beta3?

[hgsfghs]

It does, thanks.