(Tangent alert) Why does modern Firefox suck so much?

[the_r3dacted]

Um, are you asking me to download a HUNDRED different nightly builds and test each and every HUNDRED ???
Well, in a nutshell - NO!

Narrow that list of HUNDRED (no, I didn't count, but those FTPs are GIGANTIC) down to ten or so and SURE, yeah, I can do that.

No, that's inefficient. Take the first 149 build and the last 149 build to get a baseline. Extract them, delete the update.exe and pingsender.exe so it doesn't do stupid shit. Run the browser from a .bat file containing `firefox.exe -profile Profile` so it uses a self contained profile, and test. After that baseline, go somewhere in the middle, repeat. If faster, skip a bit later, test. If slower skip a bit before, test. Narrow it down that way.

[Duke]

I have too many streaming apps that don't work on Ungoogled Chromium.

Interesting. Which ones ?

These are the "font fixes" I use in Chromium/Chrome/Edge

Please can you check these pages with Chromium:
https://www.unicode.org/Public/emoji/latest/emoji-test.txt
https://en.wikibooks.org/wiki/Unicode/Character_reference/1F000-1FFFF

[The-10-Pen]

So just from those two, the very first 149 was SLOWER than 148.
And every 149 since that first just got SLOWER.
I'll grab one "in the middle" (Feb 2nd) here shortly.

[Duke]

uMatrix is not the same sort of content blocker as uBlock Origin. They're kinda made to be used together, which I do.

You may also want to check this one (remember RequestPolicy):
https://addons.mozilla.org/en-US/firefox/addon/tprb/

[The-10-Pen]

first = 452
middle = 416
last = 407

That seems like enough data points to me. 149 just got slower and slower and slower with each and every nightly.

2026-04-18_9-50-13.jpg (167.29 KiB) Viewed 161 times

[the_r3dacted]

So just from those two, the very first 149 was SLOWER than 148.
And every 149 since that first just got SLOWER.
I'll grab one "in the middle" (Feb 2nd) here shortly.

Yeah, the nightly builds are built with less optimizations (afaik could be wrong) as building with LTO and PGO and such increases the amount of time required to build. So for something they need to compile 1-2 times per day, I could see them wanting to skip that process.

It could also be that the time from last nightly to first release (I think like 1-2 months) resulted in more optimizations being added or performance reducing bugs being fixed.

uMatrix is not the same sort of content blocker as uBlock Origin. They're kinda made to be used together, which I do.

You may also want to check this one (remember RequestPolicy):
https://addons.mozilla.org/en-US/firefox/addon/tprb/

I don't think I've ever used RequestPolicy, and I'm not sure how this would be better. It looks like it does the same thing as uMatrix but with less fine grained control.

first = 452
middle = 416
last = 407

That seems like enough data points to me. 149 just got slower and slower and slower with each and every nightly.

416 to 407 is small enough that it could just be run variance. The +- is literally 39.

[The-10-Pen]

Please can you check these pages with Chromium:
https://www.unicode.org/Public/emoji/latest/emoji-test.txt
https://en.wikibooks.org/wiki/Unicode/Character_reference/1F000-1FFFF

You're on a different train of thought.
UCyborg and I have a history of discussing Chrome fonts so we just picked up where we last left off (I add this part to not end in a preposition).
We're talking about how "thin and faded" some fonts are on Chrome, we're not talking about fonts that cannot render.


Here is a 2x zoom-in *without* my font fixes:

2026-04-18_10-00-48.jpg (45.48 KiB) Viewed 150 times

Here is a 2x zoom-in *with* my font fixes:

2026-04-18_10-02-47.jpg (46.21 KiB) Viewed 150 times

[Duke]

You're on a different train of thought.
We're talking about how "thin and faded" some fonts are on Chrome, we're not talking about fonts that cannot render.

Yes, I know. I just jumped on the train because it reminded me about Chromium having troubles to display some emojis, at least on Windows 8.1.
So I just want to know how it behaves for you on Windows 10, if possible.

[The-10-Pen]

Well, on the unicode.org emoji test, the best way to "quantify" is a ctrl-f search and let it count the results.

fully-qualified = 3,948 with 3 being in the top paragraph and not really within the font subgroups
minimally-qualified = 1,033 with 3 being in top paragraph
unqualified = 247 with 3 being in top paragraph

[The-10-Pen]

That second web site has no way for me to report a quantified result.
Aside from me manually counting "squares" representing an un-rendered font (which I'm not going to count, sorry).

[The-10-Pen]

416 to 407 is small enough that it could just be run variance. The +- is literally 39.

While statistically true, I DISAGREE, v149 just got SLOWER and SLOWER with each nighly build.

Here's why I *DISAGREE*.
Multiple runs "cluster" and statistically indicate an accurate clustering.
In gun-shooting speak, the scores hit the nose each and every time, none of the bullets hit the chin, forehead, or ears.
A very tight CLUSTERING.

I'll redownload and test again. But again, the clustering is all on the NOSE.
That run variance is NOT what you think it is.

[The-10-Pen]

This is going to be my last run (at least for today).
I kind of have to feel like you are "making excuses" for v149, sorry, that's just how it sounds TO ME, lol.


Jan 12 v149 multiple Speedometer 2.1 scoring
1) 448 +/- 18
2) 457 +/- 7.1
3) 453 +/- 14
4) 455 +/- 14
5) 452 +/- 11
AVERAGE: 453

Feb 2 v149 multiple Speedometer 2.1 scoring
1) 448 +/- 23
2) 457 +/- 8.1
3) 448 +/- 7.2
4) 454 +/- 12
5) 444 +/- 12
AVERAGE: 450.5

Feb 23 v149 multiple Speedometer 2.1 scoring
1) 446 +/- 21
2) 450 +/- 8.9
3) 451 +/- 12
4) 448 +/- 10
5) 441 +/- 15
AVERAGE: 447.2


Q.E.D.
v149 just got slower and slower and slower with each nightly.
Sure, not "night and day" speed differences, but slower is slower is slower is slower.
If you want a "faster 'fox", STICK WITH v148 !!!

[The-10-Pen]

Disregard on v148 being faster. I'm not seeing that to be honest.
But this is also taking up way too much time.
I would have to do further studies, but it seems to me that the NIGHTLY v148 is SLOWER than the STABLE v148.

2026-04-18_11-39-37.jpg (170.96 KiB) Viewed 100 times

[the_r3dacted]

This is going to be my last run (at least for today).
I kind of have to feel like you are "making excuses" for v149, sorry, that's just how it sounds TO ME, lol.

I think I was thinking at this the way I do browser issues, in that it either works or it doesn't. In which there is one specific last version where a given thing works, and one first version with it doesn't. I guess that is not the case for issues like this idk.

[The-10-Pen]

Maybe (just MAYBE!), I may have been biased by the original A.I. query.

That *and* another poster cited the same, that he noticed a speed DOWNGRADE:

Anyone else noticed how fast version 148 was only for the whole speed improvement to go away in 149?

Repost of A.I. query:

[The-10-Pen]

Oops! My v148 test still had a v149 still running. This would have clearly skewed the result.
At any rate, not testing more for now. Got real life to attend to, lol.

2026-04-18_12-03-32.jpg (66.57 KiB) Viewed 78 times

[Duke]

Well, on the unicode.org emoji test, the best way to "quantify" is a ctrl-f search and let it count the results.
fully-qualified
minimally-qualified
unqualified

This is about the Unicode codes that are being used. They have evolved over time. Some old codes are now unqualified and replaced by new ones.
But that doesn't tell me which emojis are displayed and which are not.

That second web site has no way for me to report a quantified result.
Aside from me manually counting "squares" representing an un-rendered font (which I'm not going to count, sorry).

I'm not asking you to count one by one. But in both pages emojis are sorted by sections or subgroups.
Just tell me which of these subgroups are displaying squares instead of emojis. Or make screen captures if you can.

[The-10-Pen]

Seems like a sh#tload to screencap.
And the first "video cap" exceeded the attachment size.
I'll see what I can do... But holy h#ll, how would you screencap this sh#t?

[The-10-Pen]

You'll get no screencaps from me.
That first emoji test page is ONE HUNDRED AND THIRTY "pages"
Screencap #1. Page down. Screencap #2. Page down. Sorry, but F Dat.

I can capture as a scrolling video, but I get an "http error" from Board.Eclipse when trying to attach. So again, F Dat.