My Browser Builds (Part 5)

[modnar]

On 5/23/2026 at 4:08 PM, roytam1 said:

most suspected code that may cause this is GC related one. https://repo.palemoon.org/MoonchildProductions/UXP/pulls/3093

With current defaults memory consumption by the basilisk.exe will go to even north of 1600K (which never went before) and CPU constantly at 25% (one whole core occupied) and browser completely unresponsive.

I have observed that setting javascript.options.gc_on_memory_pressure to true ameliorates the memory problem (memory consumption stays at around 1000K), but no the CPU-one (I tried different settings, but haven't found one that worked).

P.S.: This memory saving setting (javascript.options.gc_on_memory_pressure) is "true" by default in newer firefox versions (or so I have read).

Edited May 25 by modnar
P.S.

[NotHereToPlayGames]

2 hours ago, AstroSkipper said:

The issue at hand goes far beyond MSFN.

Perhaps.

NOT knowing the details of how all this works and unfolded, all I can see is that it TOOK TWO TO TANGO.

ONE person did the hacking/spamming/shenanigans.  And ONE website owner that was unable to prevent it.

So let's just start picking them all off one by one.  The hacker WON.  The website owner LOST.  So let's tell tell the hacker what forum to take down next.

[AstroSkipper]

13 minutes ago, NotHereToPlayGames said:

Perhaps.

NOT knowing the details of how all this works and unfolded, all I can see is that it TOOK TWO TO TANGO.

ONE person did the hacking/spamming/shenanigans.  And ONE website owner that was unable to prevent it.

So let's just start picking them all off one by one.  The hacker WON.  The website owner LOST.  So let's tell tell the hacker what forum to take down next.

Your view of things is extremely superficial and lacks the necessary depth. Your analysis is a perfect reflection of the prevailing trend in your circles. You’d be better off asking yourself where Google, Microsoft and the SEO industry are based. That would get us to the bottom of the matter much more quickly. And rubbish simply generates more rubbish. But why am I telling you this? You must know that far better than I do. 

[NotHereToPlayGames]

35 minutes ago, AstroSkipper said:

Your view of things is extremely superficial and lacks the necessary depth. Your analysis is a perfect reflection of the prevailing trend in your circles.

Call it what you want, no skin off my back.

To me, that sounded like a violation of forum rules!

But I am MAN ENOUGH to "let it slide".

 

Bottom line, TO ME, is that I view all of this as a "the buck stops here" and I point to the FORUM OWNER as where that "buck stops".

I really do, as superficial and depthless as it may sound, BLAME THE FORUM OWNER and not Google or Microsoft or Bing or even AstroSkipper.

The buck stops at THE FORUM OWNER, nothing more, nothing less, to this "superficial and shallow" BYSTANDER.

And yeah, again, I'll let that one slide, but it really is walking RIGHT UP to the "line in the sand" of VIOLATING forum rules.

 

In the end, it's not gonna matter, we may be at the very initial stages of MSFN *ceasing to exist*.

But what do I know, I'm just superficial and shallow.  Pfft.

[AstroSkipper]

And now back to New Moon and Serpent. I’ve removed all search engines and installed Brave as well as Mojeek instead. Brave is now my default search engine, and its AI can be disabled in the Brave settings. This is particularly ideal for older hardware, as it makes the search engine much more responsive and faster. 

[roytam1]

upstream backed-out that GC change: https://repo.palemoon.org/MoonchildProductions/UXP/issues/3092

[modnar]

7 hours ago, roytam1 said:

upstream backed-out that GC change: https://repo.palemoon.org/MoonchildProductions/UXP/issues/3092

Good too; instead of collecting garbage, it's a dumpster fire itself.

P.S.: Is there a way one can disable it till saturday? 

Edited May 25 by modnar

[Guest raddy]

On 5/24/2026 at 1:34 PM, roytam1 said:

make sure your browser sends referer.

It doesn't matter, not enough RAM.., When launching

palemoon-28.10.7a1.win32-git-20260516-d849524bd-uxp-80c31f4b01-xpmod Private bytes 221MB Virtual set 507MB

palemoon-28.10.7a1.win32-git-20260523-d849524bd-uxp-829418a939-xpmod Private bytes 327MB Virtual set 604MB.

[Guest raddy]

23 hours ago, modnar said:

I have observed that setting javascript.options.gc_on_memory_pressure to true ameliorates the memory problem (memory consumption stays at around 1000K),

Interesting, but option in the current New Moon does not affect memory usage., in the previous version, it requires 15MB less at startup.

[modnar]

On 5/25/2026 at 7:06 PM, raddy said:

Interesting, but option in the current New Moon does not affect memory usage., in the previous version, it requires 15MB less at startup.

Yeah, upon further testing (today) I too observed that no setting fixes the dumpster fire that is the new garbage collection patch that's been added by upstream and already thrown out. So we'll just gonna have to rough it till the next release.

 

New finding: memory.low_physical_memory_threshold_mb seems to help work around the fiery dumpster on some sites; I set it to 1536. Check your "Physical memory" in "Task Manager".

P.P.S.: I finally reverted to 5-15 version. :-/

Edited May 27 by modnar
PPS

[newl71]

palemoon-28.10.7a1.win32-git-20260523

memory leak and load cpu 10-28% without activity

[modnar]

15 hours ago, newl71 said:

palemoon-28.10.7a1.win32-git-20260523

memory leak and load cpu 10-28% without activity

Revert to 5-15 version till saturday...

[roytam1]

New build of Serpent/UXP for XP!

Test binary:
Win32 https://o.rthost.win/basilisk/basilisk52-g4.8.win32-git-20260530-3219d2d-uxp-3c23f8ee4e-xpmod.7z
Win64 https://o.rthost.win/basilisk/basilisk52-g4.8.win64-git-20260530-3219d2d-uxp-3c23f8ee4e-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-20260530-3219d2d-uxp-3c23f8ee4e-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-28.10.7a1.win32-git-20260530-d849524bd-uxp-3c23f8ee4e-xpmod.7z
Win32 IA32 https://o.rthost.win/palemoon/palemoon-28.10.7a1.win32-git-20260530-d849524bd-uxp-3c23f8ee4e-xpmod-ia32.7z
Win32 SSE https://o.rthost.win/palemoon/palemoon-28.10.7a1.win32-git-20260530-d849524bd-uxp-3c23f8ee4e-xpmod-sse.7z
Win64 https://o.rthost.win/palemoon/palemoon-28.10.7a1.win64-git-20260530-d849524bd-uxp-3c23f8ee4e-xpmod.7z
Win7+ x64 AVX2 https://o.rthost.win/palemoon/palemoon-28.10.7a1.win64-git-20260530-d849524bd-uxp-3c23f8ee4e-w7plus-avx2.7z

Official UXP changes picked since my last build:
- Re-land: Fix devtools on 32-bit big endian platforms (0915736715)
- Issue #3053 - Implement CSSStyleSheet constructor (f5a84f4ad2)
- [js] Use size_t when inflating UTF8 (7bbdccd376)
- [NSS] Fix instances of softoken attributes freed after owning object. (300dd371b2)
- [NSS] Handle SEC_ASN1_NULL in sec_asn1e_contents_length. (9179441f96)
- [libjar] Check Jar entry names for nulls. (aadb6beb07)
- [NSS] Align PKCS7 digest array with digestAlgorithms. (089170ae83)
- [NSS] NSS_CMSContentInfo_SetContent: only modify cinfo if everything succeeds. (7167a8fabc)
- [NSS] Initialize src in SEC_PKCS5GetIV (07201fa1ae)
- [NSS] Avoid integer overflow when converting AVA value to hex string. (6511833580)
- Bug 2029771 - Heap use-after-free in [@ token_destructor] reading tok->pk11slot after nssToken_Destroy frees the token arena. (57bef0265c)
- Bug 2029782 - fix 8-byte over-read of AES-192 key buffer in x86 builds without USE_HW_AES. (92b3f6dd67)
- [netwerk] nsRequestObserverProxy ref cleanup. (bb2275d400)
- [netwerk] Make nsSocketTransport2::mConnectionFlags atomic. (bb85939429)
- Bug 2036905: Fix UDPSocketParent::ConnectInternal data race on mSocket. (8d4079866b)
- [netwerk] Make socket transport hold a reference to TLSServerConnectionInfo. (349f46bbb2)
- Bug 2027381 - improve error handling in SECITEM_DupArray with non-null arena. (b3f801743b)
- [NSS] Fix maxSize calculation in NSSUTIL_AddNSSFlagToModuleSpec. (c36357a954)
- Revert "Issue #3092 - Perform a minor GC on tab close" (3ae3595f6a)
- Revert "Issue #3092 - Initial idle GC implementation" (860d4457ae)
- Revert "Issue #3092 - Implement parallel sweeping and compaction tasks" (52ce9ebea5)
- Revert "Issue #3092 - Implement BackgroundFinalizeTask for parallel garbage collection finalization" (ac0510f380)
- Revert "Issue #3092 - Add new GC sweep tasks." (e95b229db8)
- Revert "Issue #3092 - Refactor WASM compilation handling" (d582d12ad8)
- Revert "Issue #3092 - Safely parallelize GC background finalization" (c144b4f403)
- [NSS] Fix use of uninitialized length after failed PK11_SignWithMechanism/SymKey. (2a26d9508b)
- Bug 2029818 - avoid refcount over-release in CERT_CertChainFromCert error path. (62d731d2c2)
- [DOM] Check values in audio resampling. (244c614a47)
- [layout] Hide accessible carets when needed. (4bf236902c)
- [libvorbis] Allocate memory with _ogg_malloc (e9c3451d54)
- No issue - Remove ISO-2022-JP from menu, overridability and detector. (c690e26c67)
- [gfx/layout] Simplify textruns (b611d80223)
- Bug 1784128 - Assert count passed to PR_Read/PR_Write in nsFileStreamBase fits INT32_MAX. (f83e05cf23)
- [DOM] Hold a strong ref to VoiceData in nsSynthVoiceRegistry::RemoveVoice. (2f73a3004b)
- [media] ffvpx patch: Fix leak in flac decoder in case of alloc failure. (9afa7a80cc)
- [DOM] Stop speech synthesis if the originating document is closed. (f041eb0607)

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

No official Basilisk changes picked since my last build.

Update Notice:
- You may delete file named icudt*.dat and icu63.dll 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.win32-20260530-40a79c75-uxp-3c23f8ee4e-xpmod.7z
BNavigator Win32 https://o.rthost.win/boc-uxp/bnavigator.win32-20260530-40a79c75-uxp-3c23f8ee4e-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.win32-20260530-id-656ea98-uxp-3c23f8ee4e-xpmod.7z
IceApe-UXP(suite) https://o.rthost.win/hbl-uxp/iceape.win32-20260530-id-656ea98-ia-c642e3c-uxp-3c23f8ee4e-xpmod.7z

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

[roytam1]

New build of post-deprecated Serpent/moebius for XP!
* Notice: This repo will not be built on regular schedule, and changes are experimental as usual.
** Current moebius patch level should be on par with 52.9, but some security patches can not be applied/ported due to source milestone differences between versions.

Test binary:
Win32 https://o.rthost.win/basilisk/basilisk55-win32-git-20260530-7189d3699-xpmod.7z
Win64 https://o.rthost.win/basilisk/basilisk55-win64-git-20260530-7189d3699-xpmod.7z

repo: https://github.com/roytam1/basilisk55

Repo changes:
- Revert "ported from UXP: Issue #3092 - Safely parallelize GC background finalization (e9826f55)" (f1d082056)
- Revert "ported from UXP: Issue #3092 - Fix unsafe GC multithreading changes (f0cba412)" (d27f358dc)
- Revert "import from UXP: Issue #3092 - Initial idle GC implementation (18ddd00a)" (5e38c6ebd)
- Revert "ported from UXP: Issue #3092 - Implement parallel sweeping and compaction tasks for improved garbage collection performance (3433d538)" (b24b0b646)
- Revert "ported from UXP: Issue #3092 - Implement BackgroundFinalizeTask for parallel garbage collection finalization (c0677633)" (ea9e25798)
- Revert "import from UXP: Issue #3092 - Add new GC sweep tasks. (47746b47)" (2332bc302)
- Revert "ported from UXP: Issue #3092 - Refactor WASM compilation handling (a7a75b78)" (6108ea66a)
- import from UXP: Re-land: Fix devtools on 32-bit big endian platforms (09157367) (50cedf4ac)
- ported from UXP: Issue #3053 - Implement CSSStyleSheet constructor (f5a84f4a) (ddeba088c)
- import from UXP: [js] Use size_t when inflating UTF8 (7bbdccd3) (02d728c6c)
- import from UXP: [NSS] Fix instances of softoken attributes freed after owning object. (300dd371) (cd4767c93)
- import from UXP: [NSS] Handle SEC_ASN1_NULL in sec_asn1e_contents_length. (9179441f) (0ac0961f0)
- import from UXP: [libjar] Check Jar entry names for nulls. (aadb6beb) (ccdbe4215)
- import from UXP: [NSS] Align PKCS7 digest array with digestAlgorithms. (089170ae) (6725ca403)
- import from UXP: [NSS] NSS_CMSContentInfo_SetContent: only modify cinfo if everything succeeds. (7167a8fa) (150f7251c)
- import from UXP: [NSS] Initialize src in SEC_PKCS5GetIV (07201fa1) (0cd1f405c)
- import from UXP: [NSS] Avoid integer overflow when converting AVA value to hex string. (65118335) (813a4128d)
- import from UXP: Bug 2029771 - Heap use-after-free in [@ token_destructor] reading tok->pk11slot after nssToken_Destroy frees the token arena. (57bef026) (160c8de65)
- import from UXP: Bug 2029782 - fix 8-byte over-read of AES-192 key buffer in x86 builds without USE_HW_AES. (92b3f6dd) (eae85f288)
- import from UXP: [netwerk] nsRequestObserverProxy ref cleanup. (bb2275d4) (13f345626)
- import from UXP: [netwerk] Make nsSocketTransport2::mConnectionFlags atomic. (bb859394) (30b1e9927)
- import from UXP: Bug 2036905: Fix UDPSocketParent::ConnectInternal data race on mSocket. (8d407986) (ed08f977e)
- import from UXP: [netwerk] Make socket transport hold a reference to TLSServerConnectionInfo. (349f46bb) (204c60674)
- import from UXP: Bug 2027381 - improve error handling in SECITEM_DupArray with non-null arena. (b3f80174) (4c0ddeea1)
- import from UXP: [NSS] Fix maxSize calculation in NSSUTIL_AddNSSFlagToModuleSpec. (c36357a9) (25cfe00f9)
- import from UXP: [NSS] Fix use of uninitialized length after failed PK11_SignWithMechanism/SymKey. (2a26d950) (8bc7cd06a)
- import from UXP: Bug 2029818 - avoid refcount over-release in CERT_CertChainFromCert error path. (62d731d2) (60c2cf380)
- import from UXP: [DOM] Check values in audio resampling. (244c614a) (fc43479c0)
- import from UXP: [layout] Hide accessible carets when needed. (4bf23690) (1ceaab06a)
- import from UXP: [libvorbis] Allocate memory with _ogg_malloc (e9c3451d) (3c719091d)
- import from UXP: No issue - Remove ISO-2022-JP from menu, overridability and detector. (c690e26c) (670720d41)
- ported from UXP: [gfx/layout] Simplify textruns (b611d802) (3dc9429a3)
- ported from UXP: Bug 1784128 - Assert count passed to PR_Read/PR_Write in nsFileStreamBase fits INT32_MAX. (f83e05cf) (62fdd7306)
- import from UXP: [DOM] Hold a strong ref to VoiceData in nsSynthVoiceRegistry::RemoveVoice. (2f73a300) (301814744)
- import from UXP: [media] ffvpx patch: Fix leak in flac decoder in case of alloc failure. (9afa7a80) (3e56e3dd7)
- import from UXP: [DOM] Stop speech synthesis if the originating document is closed. (f041eb06) (7189d3699)