Furthermore, here's my Firefox settings "fonts, advanced ...":
Of those settings, the only one that might have been me is the Times New Roman for serif(ied) font.
Hartmut
----- Original message ----- From: Hartmut W Sager hwsager@marityme.net To: MUUG - Round Table roundtable@muug.ca Subject: Re: [RndTbl] firefox font weirdness? Date: Friday, March 08, 2024 02:10
Oooh, ouch! I'm very sensitive to crap like that, and it has happened to me a few times over the years.
Right now, I'm on (Win 10) Firefox ver 123.0 (nothing after the 0), but it wants to update me, which, in view of what you said, I will try to hold off. Besides, I have about 40 open tabs with unfinished business.
I can also look up my Firefox settings on items of specific interest to you, since you can't view pre-123.0.2 settings.
My Firefox default font is Times New Roman 16. (Hmmm, did I do that some time ago? - I don't recall, and I typically use Verdana 10 or 12 whenever I don't like someone else's font choice.) I haven't yet looked into all the "inspector" stuff you quoted.
Hartmut
On Fri 08 Mar 2024 at 01:52:09 -06:00, Trevor Cordes trevor@tecnopolis.ca wrote:
_______________________________________________ Roundtable mailing list Roundtable@muug.ca https://muug.ca/mailman/listinfo/roundtable
On 2024-03-08 Hartmut W Sager wrote:
Thanks Hartmut. I went in there and mines was set to firefox default which appears to be Bitstream Vera. I changed it to Liberation and now my minimal html page looks fine again.
I wonder why Vera would look all messed up in firefox? My rpm logs show Vera was last updated Jan 1, which is like 4 reboots ago (lots of power outages lately), so that rpm is not the likely culprit.
The web forum textarea I mentioned still looks horrid, as its css is specifying DejaVu Sans Mono, and my system does indeed have that font; or maybe it's ignoring that and using the font css, not font family.
The firefox webdev inspector lets me see what css was selected for the final font-family and font choices, but it still shows the list for each. Is there a way to see what font the browser actually chose from the list? I know it'll go through each one from first to last until it finds one you have installed. But how can you tell which one it settled on?
The most annoying part isn't the font per se, it's that it's messing up the line heights, space between characters and underlines, font thickness/color, and the numbers appear to be from a completely different font (but maybe same family) from the letters!
I can't just point firefox bug report page to the forum because the textarea form is on a logged-in-only page.
...that should be simple to recreate on a public, static HTML page hosted on some webserver you manage, no? And if it's not simple, perhaps the bug has been misidentified. The rendering stack is pretty deep - there's *at least* five or six layers in between the TTF on disk and the pixels on screen, probably more, any of them could be the problem. e.g. is this a FF issue, a font issue, a libttf (or whatever is called today) issue, a Cairo issue... Is it even just a CSS interaction issue? Good luck. This harkens bank to the bad old days of crappy font rendering on Linux.
-Adam
Get Outlook for Androidhttps://aka.ms/AAb9ysg ________________________________ From: Roundtable roundtable-bounces@muug.ca on behalf of Trevor Cordes trevor@tecnopolis.ca Sent: Friday, March 8, 2024 9:11:04 PM To: Hartmut W Sager hwsager@marityme.net Cc: Continuation of Round Table discussion roundtable@muug.ca Subject: Re: [RndTbl] Fwd: Re: firefox font weirdness?
On 2024-03-08 Hartmut W Sager wrote:
Thanks Hartmut. I went in there and mines was set to firefox default which appears to be Bitstream Vera. I changed it to Liberation and now my minimal html page looks fine again.
I wonder why Vera would look all messed up in firefox? My rpm logs show Vera was last updated Jan 1, which is like 4 reboots ago (lots of power outages lately), so that rpm is not the likely culprit.
The web forum textarea I mentioned still looks horrid, as its css is specifying DejaVu Sans Mono, and my system does indeed have that font; or maybe it's ignoring that and using the font css, not font family.
The firefox webdev inspector lets me see what css was selected for the final font-family and font choices, but it still shows the list for each. Is there a way to see what font the browser actually chose from the list? I know it'll go through each one from first to last until it finds one you have installed. But how can you tell which one it settled on?
The most annoying part isn't the font per se, it's that it's messing up the line heights, space between characters and underlines, font thickness/color, and the numbers appear to be from a completely different font (but maybe same family) from the letters!
I can't just point firefox bug report page to the forum because the textarea form is on a logged-in-only page. _______________________________________________ Roundtable mailing list Roundtable@muug.ca https://muug.ca/mailman/listinfo/roundtable
On 2024-03-09 Adam Thompson wrote:
Oh no, that means work! :-)
https://tecnopolis.ca/fonttest.html
Interesting: all it takes to trigger the bug is a textarea with a font-family style. (See source)
You can test the textarea on that page, and see the results in a screenshot from my browser showing the bug.
I find it very strange that all the families specified appear to be mono, yet clearly the font selected is not mono (see the wwmmi line). It *is* the same font as is being selected on the real website, and it's almost an identical result... except for the numbers-lighter-color-than-letters which doesn't seem to be recreated with this simplified example.
The main thing here is the wrong font being used, *and* the crazy line spacing when letters are present! Insane! Looks perfectly fine on Chromium. And you should see what it does to underscored lines! Look at the spacing between the 123 and 456 vs the other lines. Those aren't extra NL's I've inserted!!
Ya, it smells like firefox, but could be other layers. Luckily I noticed it right away because I use this site every day. But I checked the other stuff you mentioned and Cairo hasn't been updated for months, and there doesn't seem to be any lib with ttf in the name anymore, so who knows. If ldd still means anything anymore, it looks like firefox is compiled with almost everything static on Fedora.
I guess with a simple test case I can submit a report to FF now... if someone can maybe test on their versions of FF and either me-too or not-me, along with their version #. I'd love to a) hear that <123 = not-me and b) 123 = me-too.
Not of very much help, but FF 123.0 Win64 renders it fine on my desktop. I don't have any UNIX(et al.) desktops at the moment. -Adam
-----Original Message----- From: Trevor Cordes trevor@tecnopolis.ca Sent: Saturday, March 9, 2024 8:29 PM To: Adam Thompson athompso@athompso.net Cc: Continuation of Round Table discussion roundtable@muug.ca; Hartmut W Sager hwsager@marityme.net Subject: Re: [RndTbl] Fwd: Re: firefox font weirdness?
On 2024-03-09 Adam Thompson wrote:
Oh no, that means work! :-)
https://tecnopolis.ca/fonttest.html
Interesting: all it takes to trigger the bug is a textarea with a font-family style. (See source)
You can test the textarea on that page, and see the results in a screenshot from my browser showing the bug.
I find it very strange that all the families specified appear to be mono, yet clearly the font selected is not mono (see the wwmmi line). It *is* the same font as is being selected on the real website, and it's almost an identical result... except for the numbers-lighter-color-than-letters which doesn't seem to be recreated with this simplified example.
The main thing here is the wrong font being used, *and* the crazy line spacing when letters are present! Insane! Looks perfectly fine on Chromium. And you should see what it does to underscored lines! Look at the spacing between the 123 and 456 vs the other lines. Those aren't extra NL's I've inserted!!
Ya, it smells like firefox, but could be other layers. Luckily I noticed it right away because I use this site every day. But I checked the other stuff you mentioned and Cairo hasn't been updated for months, and there doesn't seem to be any lib with ttf in the name anymore, so who knows. If ldd still means anything anymore, it looks like firefox is compiled with almost everything static on Fedora.
I guess with a simple test case I can submit a report to FF now... if someone can maybe test on their versions of FF and either me-too or not-me, along with their version #. I'd love to a) hear that <123 = not-me and b) 123 = me-too.
Speak of the devil... continuing through my email:
From: updates@fedoraproject.org Subject: Fedora 38 Update: firefox-123.0.1-1.fc38 Date: Sun, 10 Mar 2024 01:23:46 +0000 (UTC) Name : firefox Product : Fedora 38 Version : 123.0.1 Release : 1.fc38
So maybe this was a doh! release and they're already pushing out a quick fix? But ... uh... 123.0.1?
#rpm -qi firefox Name : firefox Version : 123.0 Release : 2.fc38 Install Date: Sun Mar 3 18:41:11 2024 Source RPM : firefox-123.0-2.fc38.src.rpm
Huh?? If 123.0.1 was just released, what the heck am I doing with a 123.0.2? I'm starting to smell a package maintainer mixup, someone pushing out a beta by accident?
Luckily I still have the release notice for 123.0.2:
From: updates@fedoraproject.org Subject: Fedora 38 Update: firefox-123.0-2.fc38 Date: Fri, 1 Mar 2024 01:40:45 +0000 (UTC) Name : firefox Product : Fedora 38 Version : 123.0 Release : 2.fc38
I may now be in a situation where dnf will not update the rpm! It usually will ignore older rpm files (by version). Ugh, there might be some manual dnf/rpm finagling coming my way.
Now I just have to wait a few hours until the latest rpm hits my mirrors...
You can always remove and reinstall the RPM using DNF, am I missing something? -Adam
-----Original Message----- From: Trevor Cordes trevor@tecnopolis.ca Sent: Saturday, March 9, 2024 8:39 PM To: Adam Thompson athompso@athompso.net Cc: Continuation of Round Table discussion roundtable@muug.ca; Hartmut W Sager hwsager@marityme.net Subject: Re: [RndTbl] Fwd: Re: firefox font weirdness?
Speak of the devil... continuing through my email:
From: updates@fedoraproject.org Subject: Fedora 38 Update: firefox-123.0.1-1.fc38 Date: Sun, 10 Mar 2024 01:23:46 +0000 (UTC) Name : firefox Product : Fedora 38 Version : 123.0.1 Release : 1.fc38
So maybe this was a doh! release and they're already pushing out a quick fix? But ... uh... 123.0.1?
#rpm -qi firefox Name : firefox Version : 123.0 Release : 2.fc38 Install Date: Sun Mar 3 18:41:11 2024 Source RPM : firefox-123.0-2.fc38.src.rpm
Huh?? If 123.0.1 was just released, what the heck am I doing with a 123.0.2? I'm starting to smell a package maintainer mixup, someone pushing out a beta by accident?
Luckily I still have the release notice for 123.0.2:
From: updates@fedoraproject.org Subject: Fedora 38 Update: firefox-123.0-2.fc38 Date: Fri, 1 Mar 2024 01:40:45 +0000 (UTC) Name : firefox Product : Fedora 38 Version : 123.0 Release : 2.fc38
I may now be in a situation where dnf will not update the rpm! It usually will ignore older rpm files (by version). Ugh, there might be some manual dnf/rpm finagling coming my way.
Now I just have to wait a few hours until the latest rpm hits my mirrors...
On Sat 09 Mar 2024 at 21:43:59 -06:00, Adam Thompson athompso@athompso.net wrote:
Yes, you are missing something; like Trevor already said, "that's work". :) :)
On 2024-03-10 Adam Thompson wrote:
You can always remove and reinstall the RPM using DNF, am I missing something? -Adam
Not quite so easy because repos only carry the latest rpm and the original as-released-with-the-original-Fedora-numbered-release rpm. Would have to go poking around koji or bodhi or whatever weird named system has all the rpms... unless I want to go all the way back to 111.0.1-1.fc38 as released by F38. Since I now know there are minimal dependencies, that's actually a viable option for a quick test.
...
But I don't have to as today the 123.0.1 showed up in Fedora mirrors and (strangely) dnf picked up on it as an update to install. (Strange, that means it's using dates, repo meta data outside the rpm, or something else instead of version numbers?)
And... problem solved! Voila. So it was just some wacky bug in a not-ready-for-release 123.0.2. These types of Fedora-isms used to bite me a few times a year, but it's been a while since the last time. It seemed to be getting better. Still is, I guess. Screwy fonts once a year I can live with, and I love a good challenge... I'm a bit disappointed there were no RHBZ hits on it though. Clearly it wasn't "just me".
Strangely enough, overnight I did another test after finding literally the only google hit talking about 123.0.2 and it said their (different) problem went away if you hosed the ff profile. So I used firejail --private to emulate that and it too solved my problem. So something unique to my profile *plus* the buggy ff version = font bug. (I did some major cleaning of my normal profile's about:config seeing if it was some weird setting unique to me causing problems, but nothing I did would fix it.)
Problem solved! MUUG for the win!
Argh! It came back! Went to my fave web page again today and the fonts in the textarea were screwed up again. So now I'm really baffled.
More digging... FF *does* give a way to see the fontname actually selected! The dev inspector has a font tab hidden in a dropdown arrow. It'll show the actual font selected for a single html element. Just what we need.
So... for the textarea in question, it shows something interesting:
"" Liberation Sans
Lohit Bengali
Lohit Bengali ""
I have no idea why it shows 3 fonts for 1 element. But "Bengali" and "Lohit" seem out of place. (Liberation is the one I manually selected as the FF default in previous steps.)
So: dnf remove lohit*
Restart FF, and boom, problem solved (again?). The problem appears to be the mixing of different fonts in a single element, which also explains why the numbers in some cases looked like a different font.
The FF font tab now shows just Liberation, which is as it should be.
I still don't know why this was unique to my main FF profile (as --private to firejail also solved it), nor why FF wants 3 fonts for 1 element, nor why it's magically picking lohit even though that's not specified anywhere (settings, html css). Also doesn't explain why it broke mid march when almost nothing but FF has been updated since then.
And why is anything Bengali being enabled in my profile??
I wonder if there's a chance there's some master Fedora font-choosing file / system that suddenly had "Lohit" appear (higher) in its list of sans-substitutes?
Oh! Also clear your font *cache*. This is supposed be automatic today, but ... maybe?
Get Outlook for Androidhttps://aka.ms/AAb9ysg ________________________________ From: Roundtable roundtable-bounces@muug.ca on behalf of Trevor Cordes trevor@tecnopolis.ca Sent: Friday, March 8, 2024 9:11:04 PM To: Hartmut W Sager hwsager@marityme.net Cc: Continuation of Round Table discussion roundtable@muug.ca Subject: Re: [RndTbl] Fwd: Re: firefox font weirdness?
On 2024-03-08 Hartmut W Sager wrote:
Thanks Hartmut. I went in there and mines was set to firefox default which appears to be Bitstream Vera. I changed it to Liberation and now my minimal html page looks fine again.
I wonder why Vera would look all messed up in firefox? My rpm logs show Vera was last updated Jan 1, which is like 4 reboots ago (lots of power outages lately), so that rpm is not the likely culprit.
The web forum textarea I mentioned still looks horrid, as its css is specifying DejaVu Sans Mono, and my system does indeed have that font; or maybe it's ignoring that and using the font css, not font family.
The firefox webdev inspector lets me see what css was selected for the final font-family and font choices, but it still shows the list for each. Is there a way to see what font the browser actually chose from the list? I know it'll go through each one from first to last until it finds one you have installed. But how can you tell which one it settled on?
The most annoying part isn't the font per se, it's that it's messing up the line heights, space between characters and underlines, font thickness/color, and the numbers appear to be from a completely different font (but maybe same family) from the letters!
I can't just point firefox bug report page to the forum because the textarea form is on a logged-in-only page. _______________________________________________ Roundtable mailing list Roundtable@muug.ca https://muug.ca/mailman/listinfo/roundtable
-
Adam Thompson -
Hartmut W Sager -
Trevor Cordes