355440 - Layout of text in HTML <input> seems very broken

Status: RESOLVED WORKSFORME

(Core :: Layout, defect)

Description

[Jonathan Watt [:jwatt]]

Text in a form input seems quite broken depending on the font-size specified. For some font-size values everything seems fine, for other values the caret falls behind the input text or progresses ahead of it too fast. Bizarrely, adding an explicit inline font-family property to the input element fixes things, even though the same family should be inherited by the input. Testcases to follow.

Comment 1

[Jonathan Watt [:jwatt]]

Attachment: input without explicit inline font-family - broken

In this testcase the caret will fall behind any text typed into the input.

Comment 2

[Jonathan Watt [:jwatt]]

Attachment: input with explicit inline font-family - not broken

This testcase is the same as the previous testcase except for the addition of an explicit, inline font-family property on the input element. This seems to correct the problem with the caret so that it is positioned correctly.

Comment 3

[Boris Zbarsky [:bzbarsky]]

I don't see a problem on Linux, with or without cairo.  Is this trunk-only, or also branch?  If the former, is this a cairo regression?  If not, when did it regress?

> even though the same family should be inherited by the input

forms.css specifies a font-family for inputs.

Comment 4

[Jonathan Watt [:jwatt]]

I see the bug with --enable-default-toolkit=windows too, so it doesn't look like it's anything to do with cairo. I'll try and narrow down a regression window.

Thanks for the forms.css tip.

Comment 5

[Jonathan Watt [:jwatt]]

So this is reproducible for me in a clobber build of a clean trunk Firefox source tree with the same options as the tinderbox, with the exception of having thebes turned _off_. I.e. using the following options:

--enable-application=browser --enable-application=browser --enable-svg --enable-canvas --enable-default-toolkit=windows --disable-tests --enable-static --disable-shared --enable-optimize --disable-debug

It turns out that the reason I was seeing it in a thebes build (the exact same options as the tinderboxes) was because I still had the third attachment to bug 337031 lying around in my tree. I discovered that none of the thebes firefox nightlies have this problem, and after cleaning up that old patch, my thebes builds don't either.

I find it strange that (a) the non-thebes builds have this problem; (b) the patch from bug 337031 causes thebes builds to have the problem too. Weird.

Comment 6

[Boris Zbarsky [:bzbarsky]]

Was this a problem in Gecko 1.8?  If not, can you narrow the regression range?

Comment 7

[Jonathan Watt [:jwatt]]

I don't see this problem in Firefox 2 rc1, so it seems not.

Do we have archives of Firefox trunk builds without thebes? If we do I'm not aware of them, but I'd be happy to test them if they exist. However, if we don't have them, I simply don't have the time to be checking out snapshots and creating untold numbers of builds to track this down. Then again, if cairo is our future and the bug is only present on trunk, do we actually care about it?

Comment 8

[Boris Zbarsky [:bzbarsky]]

> Do we have archives of Firefox trunk builds without thebes?

We sure do before thebes turn-on.  Was this a problem at that point?

I agree that debugging this via checkout-and-build should be saved for when this becomes a truly serious problem.

> Then again, if cairo is our future and the bug is only present on trunk, do
> we actually care about it?

We do if it blocks SVG work, or if the SVG code is just exposing a bug we have that we're not hitting in other codepaths as much for some reason.

Comment 9

[Jonathan Watt [:jwatt]]

Thebes was turned on on Windows on 2006-02-22 (see bug 323923). Unfortunately the build from 2006-02-20-08-trunk does not have the bug, so it looks like we have no easy way of tracking down this regression. :-(

What makes you think this might be connected with SVG? Note that SVG will only be supported in thebes builds soon (see bug 354866).

Comment 10

[Boris Zbarsky [:bzbarsky]]

> What makes you think this might be connected with SVG?

The references to bug 337031

Comment 11

[Jonathan Watt [:jwatt]]

Ah, I see. That was just an unrelated patch that I had forgotten was in my tree. Besides that, it is a broken hack and apparently simply wrong.

Comment 12

[Virginia (please send need info request to Brindusa Tot)]

Hi,

I tested this issue on Firefox Nightly 94.0a1 (2021-10-04) (64-bit) on Windows 10, using the test case from comment #1 and I was unable to notice the issue.
Since this seems to be an outdated issue I'll be marking it as Resolved - WFM , but feel free to reopen it with more information in case issue persists.

Regards,
Virginia