My Browser Builds (Part 4)

[AstroSkipper]

In my quite clean testing profile, setting both dom.webcomponents.enabled and dom.getRootNode.enabled to the value true with Palefill 1.26 enabled works on GitHub, i.e., GH timestamps are shown in New Moon 28.
Here are two screenshots taken from GitHub:

[AstroSkipper]

And here is a screenshot taken from @roytam1's commits on GitHub two hours later:

[VistaLover]

2 hours ago, AstroSkipper said:

setting both dom.webcomponents.enabled and dom.getRootNode.enabled to the value true with Palefill 1.26 enabled works on GitHub

... For starters, you shouldn't do that (when testing), as palefill conflicts/interferes with native WC:

https://forum.palemoon.org/viewtopic.php?p=237202#p237202

From Moonchild himself:

Second, I have always been talking about Serpent 52 (actually, its previous version of 2023-03-02); but just to humour you and be on the same ground as you, I did launch a quasi-fresh profile of latest NM28 [v28.10.6a1 (32-bit) (2023-03-10)], where I installed and enabled palefill-v1.26 and then restarted it, for good measure...

My GitHub testing repo this time is

https://github.com/violentmonkey/violentmonkey/commits/master

At first load, the relative timestamps do appear (in the "hours" range) - this is now with both palefill+native WC enabled, as you suggested; now try and continuously soft reload that page (by clicking the "reload current page" toolbar button); soon enough you'll get (I did, on my 5th attempt ) the missing timestamps bug:

If, like me, you spend hours inside GitHub, then you "can't miss it"; this bug exists, as I've tried in fresh profiles (mainly St52, but lately NM28, too) and combinations of:
a) native WC+SR enabled, palefill disabled (recommended by Moonchild)
b) native WC+SR enabled, palefill enabled
c) either of the two above, being c1) logged-in to GH or c2) logged out from GH

The only combination (for me) where the relative GH timestamps remain permanently fixed is 

d) native WC+SR disabled, palefill enabled

FWIW, with the above combination, the (latest) NM28 with fresh profile used to generate above screengrab consumes virtually 0 CPU when being minimised here:

Conclusion: Everyone's setup may behave slightly differently, in a stochastic fashion , and this makes troubleshooting and fixing bugs quite a daunting task on the part of the devs ...

Best regards.

Edited March 15, 2023 by VistaLover

[AstroSkipper]

13 hours ago, VistaLover said:

but just to humour you and be on the same ground as you, I did launch a quasi-fresh profile of latest NM28 [v28.10.6a1 (32-bit) (2023-03-10)], where I installed and enabled palefill-v1.26 and then restarted it, for good measure...

My GitHub testing repo this time is

https://github.com/violentmonkey/violentmonkey/commits/master

At first load, the relative timestamps do appear (in the "hours" range) - this is now with both palefill+native WC enabled, as you suggested; now try and continuously soft reload that page (by clicking the "reload current page" toolbar button); soon enough you'll get (I did, on my 5th attempt ) the missing timestamps bug:

In my quite clean testing profile, I opened your testing repo and performed continuously soft reload. After more than 20 times I stopped. In this profile, setting both dom.webcomponents.enabled and dom.getRootNode.enabled to the value true with Palefill 1.26 enabled works on your testing repo, i.e., GH timestamps are shown each time. In my main profile, unfortunately not as I already mentioned in a previous post.. However, these strange differences between my profiles keep bothering me.
And I have to correct my statement from a previous post which I already edited. You're right, setting dom.webcomponents.enabled to false does not lead to a 100% CPU utilization on GitHub (but on VirusTotal ). The site was continously loading the assets from the not visible part down below. Scrolling down to the assets normalized the CPU usage. I didn't notice that in my first test. So theoretically, I am able to set dom.webcomponents.enabled to false, but I'll leave it at its default value. In any case, I have to test all of that over a longer period of time.

Best regards

Edited March 16, 2023 by AstroSkipper
Update of content

[cmccaff1]

It's been a while since I posted here...and a while since I tested New Moon 28.
Though I'm a 360v13.5 'convert' and have moved on to that as my primary browser in XP, I wanted to use NM28 again, at least for a few days...seeing PaleFill brought up, which I have used with good results before, gave me a 'spark' to revisit this 'blast from the past'. I'm writing this from NM28 right now and can confirm both it and PaleFill are working great.

It's funny because NM28 is built on fairly modern web technologies, and has gotten some nice improvements over the years to bring it more in line with newer, more capable browsers. And yet it feels very 'old-school' now, not just because of the vintage Pale Moon UI but having so much legacy code from the Firefox 52 days.

Despite that, I am impressed to see what they can still do, and my appreciation for all of the years of fun I've had with old-school Gecko/Goanna/UXP browsers gives me the patience to deal with their flaws and limitations. We're a long ways removed from how things once were, so it's nice to have NM28 and other projects like it, to remind us of those 'glory days'. As long as possible, I hope these browsers can be kept alive. (Turn off JS, for sites that don't need it, and they still crush any modern browser in terms of speed.)

Thank you for all you've done, Roy! You are one of the legends of this community, and I thank you for continuing to do what you do.

[AstroSkipper]

9 hours ago, cmccaff1 said:

Though I'm a 360v13.5 'convert' and have moved on to that as my primary browser in XP,

My primary browsers in Windows XP are New Moon 28 and Serpent 52/55. I use 360Chrome 13.5 only for testing purpose in my Windows XP installation. It is great to have alternatives but I prefer UXP browsers under Windows XP. I am also an Android user, and there you have a pure Chrome/Chromium world offering the most recent browsers. Or I use Windows 7/10 to surf the internet with the more/most recent Chromium browsers. But no need to do that in my beloved Windows XP system, and 360Chrome 13.5 is not recent anymore, either. When starting my Windows XP system, I dive into another world and enjoy the recreation of all this Chrome/Chromium stuff.  All Chromium browsers are extremely restricted, especially in terms of extensions, custom buttons, customizations and modifications.  That's one of the reasons I love and prefer my UXP browsers New Moon and Serpent.  And to have them at all, I am very thankful to our creator @roytam1!

9 hours ago, cmccaff1 said:

I wanted to use NM28 again, ... gave me a 'spark' to revisit this 'blast from the past'.

Welcome back and of course, a good decision! 

Cheers, AstroSkipper 

Edited March 16, 2023 by AstroSkipper
Update of content

[roytam1]

Notice: ( especially @adata and others ) Please don't report bug(s) of my browsers using forum's Message function (to me) unless you're providing login credentials for reproducing the bug, or your report(s) will be ignored.

Edited March 16, 2023 by roytam1

[VistaLover]

15 hours ago, AstroSkipper said:

You're right, setting dom.webcomponents.enabled to false does not lead to a 100% CPU utilization on GitHub (but on VirusTotal ). The site was continuously loading the assets from the not visible part down below. Scrolling down to the assets normalized the CPU usage. I didn't notice that in my first test.

... Another villain is below "Notifications wizard":

to be found in the right GH "sidebar"... It's akin to the "Assets wizard" under GH releases page that you described; when you load a GH page containing either one, the wizards display (in the background) a constantly spinning circle, which consumes many CPU cycles ; the wizards don't display fully until you scroll past them (downwards or upwards), thereafter the CPU consumption drops (until you have to reload the page, that is ...) - this behaviour doesn't change whether you use the native WC implementation (palefill disabled) or palefill standalone (native WC disabled) ... 

NB: The Notifications wizard pictured above does not display at all when you browse GH being logged-out; the "Assets wizard", however, will still cause the heavy CPU usage until scrolled past...
As one would expect , I haven't experienced such heavy CPU usage when using 360EEv13.x to browse GH pages containing those two wizards ...

Regards.

[AstroSkipper]

3 hours ago, VistaLover said:

As one would expect , I haven't experienced such heavy CPU usage when using 360EEv13.x to browse GH pages containing those two wizards ...

When using my Android tablet, and I use it very often (as for example in this moment), I haven't experienced such heavy CPU usage in Opera, Kiwi, Brave and Firefox, either. 

Edited March 16, 2023 by AstroSkipper

[luweitest]

Edge browser download page:

https://www.microsoft.com/en-us/edge/download?form=MA13FJ

do not work in Serpent 52

[RamonUn]

On 3/16/2023 at 12:01 AM, VistaLover said:

On 3/15/2023 at 9:59 PM, AstroSkipper said:

setting both dom.webcomponents.enabled and dom.getRootNode.enabled to the value true with Palefill 1.26 enabled works on GitHub

... For starters, you shouldn't do that (when testing), as palefill conflicts/interferes with native WC:

https://forum.palemoon.org/viewtopic.php?p=237202#p237202

 

This is true if the point is to actually test the new palemoon but in theory palefill will not polyfill anything that is already implemented by the browser. So you can safely use palefill with webcompobnents enabled and benefit from the extra polyfills. Some things on GH still need extra polyfilling to work properly for example if you have no palefill you wont get the PerformanceObserver ad thus no auto completion after typing # or @. So there are still some benefits with palefill, but they are not major and they cost significant performances. hopefully things will stabilize quickly but even then palefill will still be useful.

[VistaLover]

6 hours ago, luweitest said:

Edge browser download page:

https://www.microsoft.com/en-us/edge/download?form=MA13FJ

does not work in Serpent 52

... Can confirm ; the page does load, but clicking any of the download options (buttons) doesn't initiate a download ; it's JS related:

What's worse, I couldn't get a download to happen with any of 360EEv12/13/13.5 and KafanMiniBrowser, so I'm not able to fetch the Edge files under this Vista SP2 x86 laptop; any working workaround will be appreciated ...

[UCyborg]

On 3/15/2023 at 5:01 PM, VistaLover said:

Can someone with access to "upstream" (e.g. @UCyborg ) relay this bug to them, so, perhaps, it could be investigated and, hopefully, remedied?

They fixed it in their code tree before you posted about it here. I don't know about that notification thingie causing high CPU usage, either it doesn't happen here (Pale Moon beta) or I'm missing something.

 

On 3/15/2023 at 3:46 PM, dmiranda said:

Hi @UCyborg, somewhere in this thread (can't find it) you mentioned you delete most of the root files in basilisk (except xul) as the browser can find the included in root somewhere else in the system. Can you give more details on this?  Thanks!

 

On 3/13/2023 at 12:55 AM, VistaLover said:

... What I still don't get (being a non-coder) is why disabling JS works around it - and any answer to my uBO-related "plea-for-help", please  ?

https://support.microsoft.com/...No idea, but since I don't like messing with these things unless it's the last resort, I just added the rule to my Proxomitron to rewrite links to CSS in supported fashion...

Name = "Microsoft Support - no stylesheet preload"
Active = TRUE
URL = "support.microsoft.com/ $TYPE(htm)"
Limit = 256
Match = "<link rel="preload" as="style" onload="this.rel=\'stylesheet\'" \0>"
Replace = "<link rel="stylesheet" \0>"

Just did it the way I know it can be done...

 

On 3/15/2023 at 9:59 PM, AstroSkipper said:

In my quite clean testing profile, setting both dom.webcomponents.enabled and dom.getRootNode.enabled to the value true with Palefill 1.26 enabled works on GitHub, i.e., GH timestamps are shown in New Moon 28.

On 3/16/2023 at 12:01 AM, VistaLover said:

... For starters, you shouldn't do that (when testing), as palefill conflicts/interferes with native WC:

https://forum.palemoon.org/viewtopic.php?p=237202#p237202

13 hours ago, RamonUn said:

This is true if the point is to actually test the new palemoon but in theory palefill will not polyfill anything that is already implemented by the browser. So you can safely use palefill with webcompobnents enabled and benefit from the extra polyfills. Some things on GH still need extra polyfilling to work properly for example if you have no palefill you wont get the PerformanceObserver ad thus no auto completion after typing # or @. So there are still some benefits with palefill, but they are not major and they cost significant performances. hopefully things will stabilize quickly but even then palefill will still be useful.

Palefill as provided on martok's GitHub page will indeed still impact GitHub in a significant way even when Web Components are provided by the browser because some checks aren't there so some stuff still gets injected unconditionally, but even when certain injected code has checks to not do anything if not necessary, I suppose its code must still be parsed by the browser.

I use modified version that more thoroughly checks browser features and prevents processing rules that polyfill functions that already exist in the browser. But if one must toggle web components between on and off, changes take effect after disabling/enabling extension in the new browser window, which you must open beforehand, or restart the browser.

palefill-1.25.4.xpi

Edited March 17, 2023 by UCyborg

[roytam1]

New build of Serpent/UXP for XP!

Test binary:
Win32 https://o.rthost.win/basilisk/basilisk52-g4.8.win32-git-20230318-3219d2d-uxp-85f6a4929-xpmod.7z
Win64 https://o.rthost.win/basilisk/basilisk52-g4.8.win64-git-20230318-3219d2d-uxp-85f6a4929-xpmod.7z

source code that is comparable to my current working tree is available here: https://github.com/roytam1/UXP/commits/custom

IA32 Win32 https://o.rthost.win/basilisk/basilisk52-g4.8.win32-git-20230318-3219d2d-uxp-85f6a4929-xpmod-ia32.7z

source code that is comparable to my current working tree is available here: https://github.com/roytam1/UXP/commits/ia32

NM28XP build:
Win32 https://o.rthost.win/palemoon/palemoon-28.10.6a1.win32-git-20230318-d849524bd-uxp-85f6a4929-xpmod.7z
Win32 IA32 https://o.rthost.win/palemoon/palemoon-28.10.6a1.win32-git-20230318-d849524bd-uxp-85f6a4929-xpmod-ia32.7z
Win32 SSE https://o.rthost.win/palemoon/palemoon-28.10.6a1.win32-git-20230318-d849524bd-uxp-85f6a4929-xpmod-sse.7z
Win64 https://o.rthost.win/palemoon/palemoon-28.10.6a1.win64-git-20230318-d849524bd-uxp-85f6a4929-xpmod.7z

Official UXP changes picked since my last build:
- Issue #2135 - Follow-up: Check for aElement not being null (2249fbdf4)
- Issue #2137 - Part 1: Modify :not() selector to accept a complex selector list (82fa9fb80)
- Issue #2137 - Part 2: Implement SelectorParsingFlags and use it to pass info around (3bb3c193d)
- Issue #2137 - Part 3: Don't always use the internal pseudo-class for handling negations (b257a71cc)
- Issue #2137 - Part 4: Fix namespace regression (ef36c5659)
- Issue #2135 - Destroy the host frame and restyle when there are content changes (18e693464)
- Issue #2078 - Follow-up: Ensure empty selector lists aren't iterated when serialized (2be0261d5)
- Issue #2046 - Follow-up: Fix deprot in RelativeTimeFormat. (77b618055)
- Issue #80 - Fix deprot in js/ipc. (261dc37ad)
- Issue #80 - Fix deprot in js/xpconnect. (0b039ff7a)
- Issue #80 - Move implementation of CPOWTimer::CPOWTimer to header. (c2c95eff9)
- No issue - Make content-type on JAR channels behave the same as HTTP channels (091a2aab8)
- Issue #2046 - Follow-up: Fix deprot in builtin/intl (a2d6260cd)
- Issue #2160 - Initial support for notarizing during DMG package. Added --with-macbundle-idenity configure option to set a codesign identity. If no identity is set or cross-compiling from Linux no codesigning will be done. Currently doing a full deep bundle v2 sign, instead of limited v1. (3d2a4f406)
- Issue #2148 - Don't null out RegExpObject -> RegExpShared pointer on GC. (ad8d9639a)
- Issue #2148 - Discard RegExpShared data tables when discarding regexp JIT code. (cdd91edd7)
- Issue #2148 - Make Vector not use AlignedStorage for its inline element storage (0365f940f)
- Issue #2148 - Shrink Vector from (usually) four pointers in size to three when no inline storage is used. (481069044)

No official Pale-Moon changes picked since my last build.

Official Basilisk changes picked since my last build:
- [Basilisk] Add useragent override for instagram port of MoonchildProductions/Pale-Moon commit d641959f15cb1172f6f39130b579ccf6111a6f2f and 6d41a6139440a9c396308deca1725b0db823e127 (76cd3dfac)
- [Basilisk] Port Pale Moon version of Tab fading Backport of roytam1/UXP commit https://github.com/roytam1/UXP/commit/2a32b18f465a8d068d1b1866d027374b4706cd5a.patch (4abf53728)
- [Basilisk] add tab caption fading to CtrlTab backport of roytam1/UXP commit add tab caption fading to CtrlTab (78bb7159e)

My changes since my last build:
- Reverting rev b7e45308 and 9824659d again as I can still get a crash in `RegExpShared::finalize(FreeOp* fop)` (bb547a1b5)
- Revert "Reverting rev b7e45308 and 9824659d again as I can still get a crash in `RegExpShared::finalize(FreeOp* fop)`" (0bfc6e378)

Update Notice:
- You may delete file named icudt58l.dat inside program folder when updating from old releases.

* Notice: From now on, UXP rev will point to `custom` branch of my UXP repo instead of MCP UXP repo, while "official UXP changes" shows only `tracking` branch changes.

[roytam1]

New build of BOC/UXP for XP!

Test binary:
MailNews Win32 https://o.rthost.win/boc-uxp/mailnews.win32-20230318-409cc309-uxp-85f6a4929-xpmod.7z
BNavigator Win32 https://o.rthost.win/boc-uxp/bnavigator.win32-20230318-409cc309-uxp-85f6a4929-xpmod.7z

source repo (excluding UXP): https://github.com/roytam1/boc-uxp/tree/custom

* Notice: the profile prefix (i.e. parent folder names) are also changed since 2020-08-15 build, you may rename their names before using new binaries when updating from builds before 2020-08-15.

--

New build of HBL-UXP for XP!

Test binary:
IceDove-UXP(mail) https://o.rthost.win/hbl-uxp/icedove.win32-20230318-id-656ea98-uxp-85f6a4929-xpmod.7z
IceApe-UXP(suite) https://o.rthost.win/hbl-uxp/iceape.win32-20230318-id-656ea98-ia-93af9a0-uxp-85f6a4929-xpmod.7z

source repo (excluding UXP):
https://github.com/roytam1/icedove-uxp/tree/winbuild
https://github.com/roytam1/iceape-uxp/tree/winbuild

for UXP changes please see above.