Interest in 32-Bit Linux Builds

[Basilisk-Dev]

Is there any interest in 32-bit Linux builds of Basilisk?

[frostknight]

Is there any interest in 32-bit Linux builds of Basilisk?

There might be, but if there isn't, keep it possible to make 32 bit builds, but not outright support it.

Sometimes the people who want it, aren't aware that its being offered.

Go figure, right?

[Night Wing]

I voted "No" on the 32-bit linux builds for one reason. All four of my computers (2 desktop towers, 2 laptops) have 64-bit motherboards in them.

[athenian200]

In the poll, I voted no opinion, but I thought I would still share my reasoning on this.

I don't really see any reason why there shouldn't be 32-bit builds for Linux, but it is worth noting that with GTK2 also supported, that would essentially obligate you to release four builds of Basilisk with every new release, which seems a little impractical to me. I personally see 32-bit as more valuable than GTK2, since someone running that likely is stuck on older hardware, but GTK2 is something that the people seem like they would get more upset about losing, so I can see the logic both ways.

My thinking is that an easy way to split the difference would be to just have a 64-bit GTK3 version for modern machines, and a 32-bit GTK2 version for plugin compatibility and old machines, since both 32-bit and GTK2 are more on the obsolete side of things. That is, it seems like if this is done, you should try to find a way to avoid having to do four different builds just for Linux to save yourself a bit of work.

I voted "No" on the 32-bit linux builds for one reason. All four of my computers (2 desktop towers, 2 laptops) have 64-bit motherboards in them.

Off-topic:
Umm... I know I'm being a bit of a pedant here, but technically there is no such thing as a 64-bit motherboard. Intel 64 and AMD64 are processor features, and don't really depend on the motherboard at all. While you wouldn't see it a lot in consumer products, there's nothing technically preventing a motherboard from working with both 32-bit and 64-bit processors. In fact, DDR1 RAM was 64-bit before CPUs went 64-bit, which is to say that different buses on a motherboard can be different sizes in width. 16-bit PCI and ISA slots were still usable on 32-bit and in some cases even early 64-bit systems, and there were even 8-bit ISA slots on some earlier 32-bit systems. I think in the early 1990s, there were even these Intel Inboard cards that could upgrade a machine that originally came with a 8 or 16-bit CPU to a 32-bit 386, but it wasn't ideal because the bus width was limited.

But overall, the point is that if you look deeper into motherboard design, the more you'll see that it's actually a mess of different bus widths and isn't a neat uniform 64-bits across the board. I mean, your logic isn't totally wrong, but the emphasis should be on the CPU/processor, not the entire motherboard. You should say that you have a 64-bit processor, a 64-bit CPU if you want to be technical. Most people would just say they have a 64-bit machine or a 64-bit computer... but motherboard is a specific enough term that someone who knows very little about tech will think you're very clever, and someone who knows more than you will be awkwardly deciding whether to correct you or not. When talking about different types of motherboards, the main way we talk about them is usually something like form factor (usually ATX), or the processor socket supported (something like Socket 7, AM4, or LGA1150), or in some cases the chipset (something like Z590 or B560), but 64-bit wouldn't be a useful thing to say about a motherboard.

[Basilisk-Dev]

There might be, but if there isn't, keep it possible to make 32 bit builds, but not outright support it.

The ironic part of this is that from my understanding, linking libxul requires so much RAM that you can only compile a 32-bit build of Basilisk or Pale Moon on a 64-bit kernel.

I don't really see any reason why there shouldn't be 32-bit builds for Linux, but it is worth noting that with GTK2 also supported, that would essentially obligate you to release four builds of Basilisk with every new release, which seems a little impractical to me. I personally see 32-bit as more valuable than GTK2, since someone running that likely is stuck on older hardware, but GTK2 is something that the people seem like they would get more upset about losing, so I can see the logic both ways.

In my case I was already internally building 64-bit GTK2 builds even before releasing them to the public as I hate the GTK3 scrollbars but don't want to change it in the Basilisk code in case anyone else prefers the GTK3 scroll bars. So for now at least I'll definitely keep GTK2 64-bit builds.

I wonder if the people who are using 32-bit builds would even want GTK3 32-bit builds. Perhaps maybe we should release only GTK2 32-bit?

Long term I plan to release the following builds:
OS/2
Haiku OS
MS-DOS
Windows 3.11
Unixware
OpenBSD
Linux kernel 2.4
Windows 98
Windows XP
Linux musl

[Moonchild]

The ironic part of this is that from my understanding, linking libxul requires so much RAM that you can only compile a 32-bit build of Basilisk or Pale Moon on a 64-bit kernel.

That's only the case if you use link-time code generation / whole program optimization. If you compile normally (i.e. creating final objects and then linking them together) then the linking process takes very little RAM

I wonder if the people who are using 32-bit builds would even want GTK3 32-bit builds. Perhaps maybe we should release only GTK2 32-bit?

Since we're talking Linux, you can pretty much guarantee that people will be asking for any possible combination imaginable.

[fatboy]

I agree with Moonchild, I love Linux, but it can be a can of worms.

I guess your user base should determine your builds, then keep it minimal.

I would say GTK3 - 64bit, GTK2 - 32bit for Linux is good, even though I have Basilisk GTK2 64bit installed haha

[Basilisk-Dev]

I would say GTK3 - 64bit, GTK2 - 32bit for Linux is good, even though I have Basilisk GTK2 64bit installed haha

As mentioned earlier, I won't be dropping GTK2 64-bit builds since that is what I personally am using.

[suzyne]

Off-topic:

Long term I plan to release the following builds:
OS/2
Haiku OS
MS-DOS
Windows 3.11
Unixware
OpenBSD
Linux kernel 2.4
Windows 98
Windows XP
Linux musl

I think if you could also get it running on the original Xbox that would be awesome.

[Basilisk-Dev]

Off-topic:

I think if you could also get it running on the original Xbox that would be awesome.

I know you're kidding, but I think this is actually possible since it is possible to run 32-bit x86 Linux on a modified original Xbox. You could run 32-bit Basilisk inside of that Linux OS.

[fatboy]

As mentioned earlier, I won't be dropping GTK2 64-bit builds since that is what I personally am using.

Sounds great, thank you for supporting so many formats.

[frostknight]

I voted yes, not because I plan to use it, but there might be people out there who would want to use it.

This being said, msdos, windows 3.11 windows 98 are beyond deprecated, they are mega, mega old.

I was curious if you do plan to support them or if that was a gag.

I am genuinely serious, it sounds very bizarre.

Can a windows 3.11 computer even support a web browser to connect to the internet?

I would be stunned.

OpenBSD, Linux kernel 2.4, haiku OS, linux musl, even windows xp, make sense to me, the rest sounds beyond weird.

I think you'd find more interest for 32 bit linux then the really old beyond deprecated ones.

That being said, it is up to you. As for unixware and OS/2
I don't know if those are still worth supporting, you might know though.

[athenian200]

Can a windows 3.11 computer even support a web browser to connect to the internet?

Off-topic:
Technically, yes. At least, the Windows for Workgroups version does support a TCP/IP stack and a wide variety of 10mbps ISA Ethernet cards. The last browsers supported for it are IE5 and Netscape Navigator 4.x. As long as you don't need HTTPS or a newer HTML version, I think it sort of works. I don't think any version of the open-source Mozilla suite ever supported it, but I could be wrong. The thing is, Windows 3.11 does support a limited subset of Win32, called Win32S. In theory, if you can write a Win32 application that works on Win32s and doesn't contain any features that more modern versions of Windows would have deprecated, you can create a binary that runs on Windows 3.11 all the way up to modern Windows. Probably this wouldn't be realistic with something as complicated as a web browser, but with something simpler, like, say, Freecell, you could probably do it. The trick is, it has to be 32-bit code that can be thunked to 16-bits for Windows 3.11, and run in NTVDM on 64-bit versions of Windows. That's enough of a pain that no developers I know of have seriously attempted it for non-trivial programs. More seriously, even if you did release a more up-to-date browser for an old version of Windows, it definitely couldn't support modern JavaScript, and would only be able to support a subset of HTML5 anyway. Not even sure if it could handle modern SSL at reasonable speed, honestly.

[frostknight]

The trick is, it has to be 32-bit code that can be thunked to 16-bits for Windows 3.11, and run in NTVDM on 64-bit versions of Windows. That's enough of a pain that no developers I know of have seriously attempted it for non-trivial programs. More seriously, even if you did release a more up-to-date browser for an old version of Windows, it definitely couldn't support modern JavaScript, and would only be able to support a subset of HTML5 anyway. Not even sure if it could handle modern SSL at reasonable speed, honestly.

Hmm... I am surprised that sounded outlandish to me, but sure. I would imagine though it would need to be on the level of NT or preferably XP, to be able to be at any reasonable speed though.

But who knows. If basilisk dev can pull this off, he would have to be a genius beyond a genius.

[Moonchild]

So, did everyone just miss or ignore the [ sarcasm ] ?

[frostknight]

So, did everyone just miss or ignore the [ sarcasm ] ?

He didn't make enough indications that it was sarcasm. I had figured it might be, but I was confused as to why he sounded so serious in the post.

I was hoping it was sarcasm. Windows 3.11, msdos, windows 98 are all ULTRA ANCIENT LOL



Others are semi ancient and then there are ones that are actually reasonable.

Well these sounded mostly reasonable:
HaikuOS*
OpenBSD
Linux kernel 2.4
Windows XP*
Linux musl

The ones with asterisks, are only midlly strange but I could see people doing that.

What did you think I meant by gag?

xD

But yeah, I took him half seriously because I didn't see the sarcasm bar. I wonder if its the background wallpaper I have...

[frostknight]

32-bit?

Long term I plan to release the following builds:
OS/2
Haiku OS
MS-DOS
Windows 3.11
Unixware
OpenBSD
Linux kernel 2.4
Windows 98
Windows XP
Linux musl

Well played, I actually took your sarcasm as if it was possible you would.

I love the humor there. Thank God your not insane enough to do something that cuckoo... it would be a nightmarish headache for you lol.


As a side note, I didn't see the sarcasm highlight until I pretended to reply to the message and then I saw the sarcasm markings. I don't know why it was so hidden...

Weird...

Last edit:

I noticed, that only if I hover my pointer over the sarcasm, do I see:

"This is meant to be taken as sarcasm!"

Very strange...

[Basilisk-Dev]

Assuming that they didn't want to apply a million patches, if someone were to offer to compile an OpenBSD build in the same way that dbsoft does the FreeBSD builds I'd be willing to release it. That being said, OpenBSD is very unstable from the standpoint that it frequently breaks the userland ABI between versions so I don't know if that would be practical.

That being said, it is up to you. As for unixware and OS/2
I don't know if those are still worth supporting, you might know though.

There is actually a small community of people on the Web who still use OS/2. There is also a vendor who maintains an updated version OS/2 called ArcaOS. The name of the vendor is Arca Noae. They have a direct contract with IBM to update the OS, but IBM did not give them the source code so they directly edit the assembly for the OS/2 binaries. They've got it running on modern ACPI systems and they actually have it booting in UEFI instead of legacy BIOS, among with other several changes including drivers for modern hardware.

At one point they looked into porting UXP to OS/2 since they actually have a working port of Firefox 45, but sadly decided to port QTWebEngine instead.

I actually have a paid license and use it on an old ThinkPad T61 when I feel like running some old DOS or Windows 3.1 programs.

Updated OS/2 operating system: https://www.arcanoae.com/
OS/2 community site: https://www.os2world.com/cms/

But who knows. If basilisk dev can pull this off, he would have to be a genius beyond a genius.

Correct, I am a genius beyond a genius

[frostknight]

Assuming that they didn't want to apply a million patches, if someone were to offer to compile an OpenBSD build in the same way that dbsoft does the FreeBSD builds I'd be willing to release it. That being said, OpenBSD is very unstable from the standpoint that it frequently breaks the userland ABI between versions so I don't know if that would be practical.

I would say, probably it wouldn't be as unstable if you only focused on the latest OpenBSD release. Otherwise, you might very well be right.

There is actually a small community of people on the Web who still use OS/2. T

didn't know this, interesting

There is also a vendor who maintains an updated version OS/2 called ArcaOS. The name of the vendor is Arca Noae. They have a direct contract with IBM to update the OS, but IBM did not give them the source code so they directly edit the assembly for the OS/2 binaries. They've got it running on modern ACPI systems and they actually have it booting in UEFI instead of legacy BIOS, among with other several changes including drivers for modern hardware.

Pity they have to support UEFI though, that crap is a bloated mess. It would be one thing if the network stack beneath was always off unless enabled by the user, but that isn't the case.

Correct, I am a genius beyond a genius

When I said that, I meant if you could bring it to win 3.11 and msdos. Which yes, that would take an insane amount of skill.

IF you are indeed a genius beyond a genius, I am sure you could.

[Basilisk-Dev]

The poll results seem split basically even between people not wanting 32-bit builds and people who do want 32-bit builds.

While many people said no, there are several people who feel that we should introduce 32-bit builds. Due to the number of people who said yes, I do plan to start compiling and releasing 32-bit builds in the future.

It won't be in the upcoming release at the end of October, it will likely be in the November release. I need to work on updating the Docker based build environment to have all of the libraries required to build Basilisk as a 32-bit application and I won't have much time to do that between now and the end of October.