Nicholas McAnespy Posted June 3 Posted June 3 4 hours ago, roytam1 said: since VC5 and VC6 uses same runtime (MSVCRT.DLL) so it won't crash. 1.5 or later since they have real tabs, not layers. BTW I wonder if you can upload the sources of your 38ESR-VC2010? https://codeberg.org/Nicholas_McAnespy/FxVC10Mods/releases 2
Nicholas McAnespy Posted June 3 Posted June 3 (edited) Since I cannot guarantee the source code compiles properly, I should ask, does it work for you? Edited June 3 by ClassicNick
roytam1 Posted June 3 Posted June 3 5 minutes ago, ClassicNick said: Since I cannot guarantee the source code compiles properly, I should ask, does it work for you? haven't check, just imported it to a local repo with official 38.8.0esr as base to see how many differences there. may try to work on it(either try to build it and fix errors that encountered, or even put some NM27 bits on top of it to produce a pseudo-NM27) later.
roytam1 Posted June 3 Posted June 3 (edited) 10 hours ago, ClassicNick said: Since I cannot guarantee the source code compiles properly, I should ask, does it work for you? with lots of fixes, it works. changes: https://github.com/roytam1/firefox38-vc2010/compare/dd0315735742429dbaae309babccfc08583f9310...HEAD with a js rdtsc hackfix, it can work on a 486. https://x.com/roytam1/status/1797637584493101276 Edited June 3 by roytam1 1
roytam1 Posted June 4 Posted June 4 On 6/3/2024 at 4:22 AM, ClassicNick said: I'm still trying to port NSS 3.42 beta to the Mozilla/Gecko 1.8.0 codebase. When I asked you how difficult it would be, it was because I got an error "plarena.h no such file or directory". Now I'm cherry picking code from RetroZilla's nsprpub directory to solve unresolved externals caused by NSS 3.42 beta. I wonder if you can just replacing NSPR with retrozilla's. NSPR 4.7.6(4.7.7) is good enough for them.
roytam1 Posted June 4 Posted June 4 (edited) On 6/3/2024 at 9:41 PM, roytam1 said: with lots of fixes, it works. changes: https://github.com/roytam1/firefox38-vc2010/compare/dd0315735742429dbaae309babccfc08583f9310...HEAD with a js rdtsc hackfix, it can work on a 486. https://x.com/roytam1/status/1797637584493101276 trying to import NSS 3.42 Beta in this, it builds but CORS seems to be broken and it refuses to load many https sites like ddg. EDIT: the problem is not in NSS, but 38esr's HTTP/2 encoding is too old. EDIT2: changes pushed. https://github.com/roytam1/firefox38-vc2010 and works even on a 486: binaries for testing: original NSS from 38esr: http://o.rthost.win/gpc/files1.rt/firefox-38.8.0-vc10-nss319.7z updated NSS from retrozilla (TLS 1.2): http://o.rthost.win/gpc/files1.rt/firefox-38.8.0-vc10-nss321.7z updated NSS for TLS 1.3: http://o.rthost.win/gpc/files1.rt/firefox-38.8.0-vc10-tls13.7z Edited June 7 by roytam1 2
schwups Posted June 5 Posted June 5 First I tried this FF38/TLS1.,3 on my main machine Win ME - KernelEx 25/26 - Kext. Yes, indeed it starts and runs (compat. mode XP). Besides Icecat 38 I have one more FF38 fork working on ME, but of course basically with the same issues as with version 35 I documented. I am curious, because I had tried to start versions 36 - 41 myself for a long time without success. Other tests could follow on a bit different systems. The plugin container can run in the background after close. So I switched it off. It makes sense to disable Javascript (I installed QuickJava), due to its age-related infirmities. It will freeze on many sites, also on this msfn board. FF38 OT.
Nicholas McAnespy Posted June 6 Posted June 6 40 minutes ago, schwups said: First I tried this FF38/TLS1.,3 on my main machine Win ME - KernelEx 25/26 - Kext. Yes, indeed it starts and runs (compat. mode XP). Besides Icecat 38 I have one more FF38 fork working on ME, but of course basically with the same issues as with version 35 I documented. I am curious, because I had tried to start versions 36 - 41 myself for a long time without success. Other tests could follow on a bit different systems. The plugin container can run in the background after close. So I switched it off. It makes sense to disable Javascript (I installed QuickJava), due to its age-related infirmities. It will freeze on many sites, also on this msfn board. FF38 OT. As far as I know, Firefox 10 and later run very slowly on Windows 98SE/Me with KernelEx. Do you have a computer with a newer version of Windows to test it on? I think Windows XP SP2 will work with that Firefox 38.8.0esr custom build. @roytam1 Are you using unified C++ sources, or deunified C++ sources.
roytam1 Posted June 6 Posted June 6 1 hour ago, ClassicNick said: As far as I know, Firefox 10 and later run very slowly on Windows 98SE/Me with KernelEx. Do you have a computer with a newer version of Windows to test it on? I think Windows XP SP2 will work with that Firefox 38.8.0esr custom build. @roytam1 Are you using unified C++ sources, or deunified C++ sources. I haven't change any moz.build in my tree so it is the same as yours i.e. deunified.
Nicholas McAnespy Posted June 6 Posted June 6 59 minutes ago, roytam1 said: I haven't change any moz.build in my tree so it is the same as yours i.e. deunified. When I did my build, I got 7 unresolved externals after linking xul.dll (brwsrcmps.dll?), and my solution was to forcefully deunify the sources. Now I remember, because I didn't fully revert the disable unified sources bug, the sources now build deunified by default, with --enable-unified-compilation having no effect. How much time does it take to compile that Firefox 38.8.0esr VC2010 modded version on your computer? If I remember correctly, it's 56 minutes with New Moon 27.9.6 20190223, and ~2 hours with Firefox 38.8.0esr with my underclocked (50% frequency) AMD Phenom X3 8600.
roytam1 Posted June 6 Posted June 6 2 minutes ago, ClassicNick said: When I did my build, I got 7 unresolved externals after linking xul.dll (brwsrcmps.dll?), and my solution was to forcefully deunify the sources. Now I remember, because I didn't fully revert the disable unified sources bug, the sources now build deunified by default, with --enable-unified-compilation having no effect. How much time does it take to compile that Firefox 38.8.0esr VC2010 modded version on your computer? If I remember correctly, it's 56 minutes with New Moon 27.9.6 20190223, and ~2 hours with Firefox 38.8.0esr with my underclocked (50% frequency) AMD Phenom X3 8600. ~23mins full build on Ryzen 5800X and WD SN750 SSD on Win7 (with `-j14` parallel make jobs). 1
Nicholas McAnespy Posted June 6 Posted June 6 1 hour ago, roytam1 said: ~23mins full build on Ryzen 5800X and WD SN750 SSD on Win7 (with `-j14` parallel make jobs). Do you run your Ryzen 7 5800X at a lower frequency than the default speed? I'm curious now what your build times are with RetroZilla, and New Moon 27.9.6? With RetroZilla, I get 24 minutes, and possibly less with Firefox 1.5.
roytam1 Posted June 6 Posted June 6 1 hour ago, ClassicNick said: Do you run your Ryzen 7 5800X at a lower frequency than the default speed? I'm curious now what your build times are with RetroZilla, and New Moon 27.9.6? With RetroZilla, I get 24 minutes, and possibly less with Firefox 1.5. CPU setting are all default with Auto PBO, RAM is XMP-3200MHz. for NewMoon/Serpent/etc. usually needs ~12mins to rebuild here (with `-j 11`). for retrozilla, it needs ~30mins to finish, since cygwin-based build system is slow (not matter how fast compiler and CPU runs)
Nicholas McAnespy Posted June 6 Posted June 6 12 minutes ago, roytam1 said: CPU setting are all default with Auto PBO, RAM is XMP-3200MHz. for NewMoon/Serpent/etc. usually needs ~12mins to rebuild here (with `-j 11`). for retrozilla, it needs ~30mins to finish, since cygwin-based build system is slow (not matter how fast compiler and CPU runs) Tip for RetroZilla: Gecko 1.8.0b3 and later support building with MSYS, so that means Mozilla-Build will work with them. rn10950 used Mozilla-Build 1.2 to compile RetroZilla. If you want a native Windows command prompt window instead of rxvt like Mozilla-Build 1.2 uses, Mozilla-Build 1.3 is also suitable to build RetroZilla. Is it possible you are using mk_add_options MOZ_MAKE_FLAGS="-j1" simply because it's the default setting? What I notice with older Mozilla code (pre 2012) is you need a graphics adapter/card to take advantage of a powerful CPU. Wait... What version of Windows are you using to compile RetroZilla? In my experience, Windows 2000 and XP are the best choices for RetroZilla compilation.
schwups Posted June 6 Posted June 6 (edited) 6 hours ago, ClassicNick said: As far as I know, Firefox 10 and later run very slowly on Windows 98SE/Me with KernelEx. Do you have a computer with a newer version of Windows to test it on? I think Windows XP SP2 will work with that Firefox 38.8.0esr custom build. I am not aware, that versions 12-23 work usable at all with KernelEx. Those are freezing after a few seconds. With later versions, it depends on the hardware (i.e. CPU) whether Firefox is slow. With a Intel Prescott it runs noticeably slower than with a Core2Duo Wolfsdale. With the last mentioned it does not run slower than on XP or maybe minimally. There is no need to measure the loading times for comparison. That's not worth it. Yes I've XP, Vista and Win7, too. Edited June 6 by schwups
Recommended Posts
Create an account or sign in to comment
You need to be a member in order to leave a comment
Create an account
Sign up for a new account in our community. It's easy!
Register a new accountSign in
Already have an account? Sign in here.
Sign In Now