Nicholas McAnespy
Content Type
Profiles
Forums
Events
Posts posted by Nicholas McAnespy
-
-
On 5/14/2025 at 5:54 AM, roytam1 said:
as upstream moved to vc2013-unfriendly environment, no new changes can be applied to NM27 and its friends.
incoming patches are still welcomed.
Arctic-Fox now has an XML parsing error when run on Windows, but is unrelated to Visual C++ 2013/2015 compilation. If I make it work using Visual C++ 2013 update 4, would you still accept the compatibility changes?
0 -
On 7/5/2024 at 11:47 PM, roytam1 said:
finally got a working PM2796-vc10 (and also bring back webrtc, webgl, and mediasource): https://github.com/roytam1/palemoon27/commit/5d5e647b63c1b729e937a1bdd70a4fc18bff6d90
binary for testing: http://o.rthost.win/gpc/files1.rt/palemoon-27.9.6.win32.vc10.7z
and still working with my kernelxp wrapper on XP RTM:
1. I would still like to know more about the "kernelxp" wrapper, and how you're able to get around the EncodePointer, and DecodePointer errors in Windows XP.
2. Minor annoyance: Unless the Visual C++ 2005/2008 redist files are installed in Windows XP, I can't run a version of Firefox compiled with either Visual C++ 2005 or 2008 due to "This application has failed to start because the application configuration is incorrect". Setting WIN32_REDIST_DIR, or --enable-jemalloc didn't solve the problem. Do you know of a way I can build Firefox/New Moon so that I can just bundle the Visual C/C++ runtime files and have the browser load?
0 -
@roytam1 I have a Visual C++ 2013 update 4 build of your New Moon 27.10.0 20241220 build, taken from the palemoon27-master branch in your repository. Previously, you had merged Arctic-Fox commits into your palemoon27 repository, and "squashed" them so that all the Arctic-Fox commits you merged will appear as 1 commit. If I have the source code to palemoon27-master, and the Arctic-Fox-master git information (presently at version 45.2 from January 6th 2025), what git command do you recommend I use in order to merge commits from Arctic-Fox (I only know git cherry-pick)? Also, I asked you if you could create a VC2015 branch in your palemoon27 repository, and see if you could merge the latest Arctic-Fox commits to that branch. I finally decided to try a VC2015 build of New Moon 27.10.0 20241220, and it failed 8 minutes in with errors in the type_traits file built into Visual C++ 2015, but when I tried a VC2013 build, it compiled cleanly taking 76 minutes. That means creating a "VC2015" branch for the purpose of evading Visual C++ 2013 compatibility is not necessary in my opinion, but a "tracking" branch that is not meant to be compiled but merged to may be helpful to determine if merging commits from Arctic-Fox will still work (I'm scared of git).
1 -
Update: I managed to get half way through a Visual C++ 2008 compatible mod of New Moon 27.9.7, but I sometimes build with Visual C++ 2010 to see if the browser crashes or not. I managed to make tabbed web browsing impossible due to the content on the 1st tab spreading across all the tabs, and leaving only that 1st tab usable. Opening new windows for every tab still works though... I think the problem lies somewhere in the GFX directory, but I could be wrong. @roytam1 What do you recommend I do now?
Edit: I will continue with my current build to get it compiling under Visual C++ 2008, and then try fixing my problem on the next build. Also, my way of getting around the decltype function used in layout/base/nsChangeHint.h breaks tabbed browsing anyway.
0 -
@roytam1 Would it be possible to create a VC2015 branch in your palemoon27 repository, then merge the latest Arctic-Fox commits? If you can build it using Visual C++ 2015, that would be a great start so I can try my hand at a Visual C++ 2013 build.
0 -
5 hours ago, Leokids123 said:
Will you do a VC4 build for MIPS Retrozilla?
VC4 does not support the OPT:NOICF flag used in js/src/Makefile.in. As far as I remember, disabling optimization flags should work, but will mean using the resulting application will be slow. As a note, see bug 151066. It's been a while since I've tried compiling Mozilla source code using Visual C++ 4.x, but I do remember having an error in layout/style/nsRuleNode.h, but I don't remember the message or what caused it.
0 -
@Leokids123 I have the project marked as cancelled because I'm still struggling to make Firefox 3.0 work with Windows 9x. Brutal honesty: After having runtime execution failure in Firefox 3.0b5, and having flaky Windows 98 compatibility in Firefox 3.0a7, I have become discouraged and scared of continuing with this project. I would like if someone (mainly @roytam1) could upload my Firefox 3.0a4 mods from GitHub, so I can try again at Firefox 3.0a7 Windows 9x compatibility before moving toward Firefox 3.0 Windows GFX builds.
0 -
On 10/11/2024 at 6:53 AM, roytam1 said:
Notice about end of NM27 based browsers
Since AF is now importing post-Fx48 changes, which is more and more VC2013-unfriendly, making me struggling on trying to rewrite them to VC2013-compatible syntax.
But it seems that this can't last long and even it builds, it hits segfault in run time in various memory locations and it different code paths, It looks like I come to a blind alley.
And since the only reason for me to keep building because I want a VC2013 build, so this may be the time to stop here unless I can find a way to continue.
I hope I can bisect my last batch of imported commits and undoing the faulty (to me) commit and create a "last release" for AF, NM27, and KMG76.i.e. NM27 and KMG76 build to-be-released in tomorrow should contain last stable code before landing faulty components like SharedImmutableStringsCache.
EDIT: To clarify, Pull Requests of NM27 engine will still be reviewed and merged, doable fixes and patches suggested may still be applied, if any. But massive merging from ArcticFox will be stopped.
I was wondering why your GitHub account says UXP updated Y hours ago, but palemoon27 updated 2 weeks ago. I understand fully why you don't plan to release future New Moon 27 or K-Meleon 76.5 Goanna builds if it makes Visual C++ 2013 compatibility more difficult, especially seems how I never learned to use Visual C++ 2013. Confession: I'm getting crashes in mozjs.dll relating to js/public/HashTable.h in New Moon 27.9.6 when trying to use Visual C++ 2008 compatible syntax, so I think I will give up. Instead, I will opt for either a Firefox 24 (works well enough for me), or a Firefox 31 fork. Can you build K-Meleon 74 Gecko using the Mozilla 24.8.1 codebase?
0 -
On 10/16/2024 at 7:58 AM, modnar said:
You need no mobile version, just normal version with h264ify extension (in Serpent 52.9). And yeah their HD quality is not what it used to be - I believe they're just saying it's at 1080p when in reality they're doing some monkey business in the back.
I heard they reduced the bitrate for their videos. I've never used a 1080p camera, and I don't have much experience with 720p either, but a 640x480 video needs a minimum bitrate of 500 Kbps, but 1 Mbps should typically be high enough if encoded using h264, but there are circumstances when even 2 Mbps isn't very high. Videos with high motion will need a higher bitrate than videos where the camera doesn't move. A bright location will need much higher bitrate than a dark location will. I also believe targeting a constant bitrate will cause more blurriness than targeting a variable bitrate.
1 -
11 hours ago, AstroSkipper said:
A RAM consumption of round about 700 MB for only one YouTube video?
I remember when I use to be able to load YouTube in ~200 MB RAM including K-Meleon 76.1 Goanna, and ~250-400 MB after the switch to the "Polymer" UI. What caused RAM usage to increase to > 600 MB? I also noticed YouTube has much higher CPU consumption than it did 4 years ago.
1 -
@roytam1 I downloaded your "palemoon27-pm27100-vc2010" source code, and tried removing C++ 11 move/forwarding references. I now get a crash in mozjs.dll. Also, I can't do a debug build due to a "program database manager mismatch" error. If you can upload the source code so I can see my changes on GitHub, that would be great. Note: I removed the delete constructors in js/src/jsobj.h lines 544-545. I also used Visual C++ 2010 to compile this, so I assume my modifications are incomplete. https://codeberg.org/Nicholas_McAnespy/palemoon27/releases/tag/2796-MoveRef-Busted
0 -
@K4sum1 Thank you for doing public releases of lun3r and Hydra. I prefer to use a browser based on Pale Moon, and I build browsers based on Arctic-Fox/Pale Moon 27, so it's nice to have a browser that I can leave running, while testing my own browsers that share the Pale Moon profile.
1 -
@roytam1 I have been getting some errors related to "NodeInfo is not a member of mozilla::dom::mozilla" while trying to compile New Moon 27.9.7 using Visual C++ 2008. My way of solving those errors means de-unifying the C++ sources.
0 -
17 hours ago, Nicholas McAnespy said:
While you're doing that, I tried compiling the New Moon 27.9.7 VC2010 mod using Visual C++ 2008, and hit a roadblock due to compiler errors caused by C linkage where it's not appropriate for it to be specified. I do not know where the C linkage is though...
Update: Visual C++ 2008 doesn't like using <intrin.h> in nsprpub/pr/pratom.h, and prbit.h.
0 -
4 hours ago, roytam1 said:
don't know KM76 because its tree is pre-patched with custom patchset. see: https://github.com/roytam1/palemoon27/blob/pm27100-vc2010/toolkit/xre/WindowsCrtPatch.h#L10
With further testing, I noticed the older K-Meleon 76 Goanna builds (76.1.1) lack GetLogicalProcessorInformation in xul.dll, but have it in the gmp* files, libavcodec files, and MSVCR120.dll. I believe Firefox, Arctic-Fox, and Pale Moon have some sort of hack to allow execution on Windows XP SP2, even though Visual C++ 2010 and later do not naturally support it, which is probably why K-Meleon 76 Goanna doesn't work with Windows XP SP2. Since MSVCR90.dll is the last version to not call GetLogicalProcessorInformation, I hope this problem fixes itself if/when we get Tycho/Goanna3 compiling using Visual C++ 2005 and/or 2008.
0 -
On 7/23/2024 at 12:13 PM, roytam1 said:
finally got some time to download it and have a look on it. so you decided to skip the VC2010-faulty rev and move on to later revs.
meanwhile I'm still trying to merge newer revs to https://github.com/roytam1/palemoon27/tree/pm27100-vc2010
While you're doing that, I tried compiling the New Moon 27.9.7 VC2010 mod using Visual C++ 2008, and hit a roadblock due to compiler errors caused by C linkage where it's not appropriate for it to be specified. I do not know where the C linkage is though... If you can find where "GetLogicalProcessorInformation" is defined, that would be great, but could be difficult because it is called from xul.dll in K-Meleon 76 Goanna, but no such call exists in xul.dll from New Moon 27. Admittedly, I haven't tested whether xul.dll in K-Meleon 76 Gecko calls GetLogicalProcessorInformation, and thus, fails to execute in Windows XP SP2.
0 -
On Sat Jul 20 2024 (GMT+0000) at 12:34 PM, roytam1 said:
it is always better come with changelogs (even just come with git's default messages). avoiding all-changes-in-one-commit will be better for this type of operation.
The Source code containing git information is uploaded to https://github.com/ClassicNick/palemoon27/releases/tag/27.9.7-VC2010
1 -
@roytam1 I didn't bother cherry picking d918e4f. The last commit I cherry picked was 2d9c22e, which reverted e186cc7, which caused the browser to crash when opening a new tab. I'm also sitting on a local build of New Moon 27.9.7, so when I upload it, do you want the git information along with the source code, or do you just want the source code?
0 -
@roytam1 Now that I'm cherry picking commits from our palemoon27 repositories that are dated ~September 2019, I notice commit partial hash d918e4f (9/17/2019) apparently contains e10s code that you didn't like at the time (Accusation caused by f25cb74, which was dated 9/20/2019). Should I try cherry picking d918e4f, or should I move on without it? Also, I decided to read some comments on your 1st browser builds thread, and you said the Ryzen 7 3700X with your SSD caused build times to be 1/3 of what they were with your previous hardware. What were the specs of the previous computer you did your browser builds on?
0 -
30 minutes ago, roytam1 said:
I think you may still able to use non-HTTPS for uploading.
for example, one of my local repo still has this in git config:
[remote "origin"] url = git@github.com:roytam1/rtoss.git fetch = +refs/heads/*:refs/remotes/origin/*
and I can still able to `git push` from this repo.
My Git commands presently are "mkdir palemoon27-27100-VC2010-WIP" "cd /F/palemoon27-27100-VC2010-WIP" "git clone "https://github.com/ClassicNick/palemoon27.git" and Ex. "git cherry-pick 593d1c0". Basically, my method is cherry picking all commits from newer branches. That way, I don't need to worry as much about git conflicts. As for SSH, I don't know how to add an SSH key. Now that I can cherry-pick commits, after how many changes should I upload the source code?
0 -
On 7/11/2024 at 1:10 AM, roytam1 said:
pm2796-vc2012 ends after dropping in fx38vc10pm changes, while pm27100-vc2010 tries to merge newer changes before changing version to 27.10.0
git for XP should still work with current github repo.
After a day of trying out some git commands, I realized Git 2.10.0 does not support Codeberg due to a problem with the SSL certificate. I finally logged back in to my GitHub account, and also realized I can download whatever I want from Git 2.10.0, but I can't upload anything from there because password authentication is no longer supported on GitHub. What git command do you recommend using to port palemoon27-classic, and eventually palemoon27-master commits to palemoon27-pm27100-vc2010?
0 -
@roytam1
Before I do any more work on Visual C++ 2008 SP1 compatibility, should I use the pm2796-vc2012 branch, or the 27100-vc2010 branch? If it's possible to use Git 2.10.0 in Windows XP SP3, and if you have an excessive amount of patience, I would like it if you could help me use it.0 -
54 minutes ago, roytam1 said:
now there is a problem when merging 0b23a2783d03b3f5fe404544db3043689a9a91d0 and compiler doesn't like changes from Bug 1139552
In js/src/gc/Heap.h, #include "mozilla/TypedEnum.h", and then check if you can get away with using MOZ_BEGIN_ENUM_CLASS(AllocKind, uint8_t) in place of enum class AllocKind : uint8_t {. Also, replace } with MOZ_END_ENUM_CLASS(AllocKind). I do not guarantee that will work, but it typically will. If it doesn't, you'll need to use enum : uint8_t {. What error message are you getting with bug 1139552?
0 -
22 hours ago, roytam1 said:
no idea how to fix it, tree is pushed to github. https://github.com/roytam1/palemoon27/commits/pm27100-vc2010
I was hoping I would get the broken source code with bug 1134425 already applied, so I could see what removing the characters = CanGC from template <typename T, AllowGC allowGC = CanGC> would do.
0
My Browser Builds (Part 5)
in Browsers working on Older NT-Family OSes
Posted
That pull request actually has conflicts that I can't resolve using GitHub web, and I don't know how to upload anything to GitHub through git.