Glibc removes DT_HASH in favor of DT_GNU_HASH.

[athenian200]

So, this is what they did:

https://www.youtube.com/watch?v=Uj2z-3qAchA

They dropped support for DT_HASH, which is a standard part of the SysV ABI, in favor of DT_GNU_HASH, which is poorly documented and glibc-specific. This not only breaks a lot of existing applications, but it puts pressure on developers to design things according to a "standard" that only exists in glibc, potentially making it harder to develop for both Linux and alternate Unix operating systems (or even alternate libcs on Linux, for that matter). I don't think this really affects Pale Moon, but it does affect a lot of other cross-platform tools.

I would actually like to go on record as saying that glibc is a big factor among several reasons I will never go back to using Linux as a primary OS myself. I'll still help other people with it as best I can because I have the skills to deal with this insanity, but there's a reason I don't want Linux on my own system as a daily driver. SunOS has a libc that is developed alongside the kernel and tends to avoid breaking userspace and requiring constant code/build system changes the way glibc does seemingly every other week.

[jobbautista9]

It's not only the fact that DT_GNU_HASH is poorly documented that is the problem, they've not written any good documentation about it since they've introduced it in 2006! If they really wanted their own symbol lookup table to be a replacement, they should've written some documentation about it and marked DT_HASH as "deprecated" (which they didn't for 16 years and was still marked as "mandatory" until 2.36 released) in the first place. However I guess they're going to make an excuse that the downstream distros made the build config of their glibcs only include DT_GNU_HASH, thus it's "not their problem". But here's the catch: those same distros they're talking about now reverted their build config changes and made glibc include both tables again. That guy in the video is right: if all of downstream have to fix the problem, it's no longer the downstream's problem; it's the upstream's.

This is why I think Linux is never going to be a serious competitor in the desktop, if devs involved in the core libraries still think they can recklessly break backwards compatibility. Which is sad, because I love Linux.

[athenian200]

This is why I think Linux is never going to be a serious competitor in the desktop, if devs involved in the core libraries still think they can recklessly break backwards compatibility. Which is sad, because I love Linux.

Yeah. I actually think the dependency on GNU is holding Linux back. The most successful OS based on a Linux kernel is, unfortunately, Android. Which uses Bionic as a libc and replaces pretty much the whole GNU stack. The problem is that I find a lot of GNU devs think everyone should be building open source applications from source anyway, and if they are relying on binaries that were built by someone else and don't have source code or don't know how to rebuild them, then they "deserve what they get." I always kind of thought GNU deliberately likes to break binary backwards compatibility as a way of pushing the open source philosophy and putting pressure on people to only rely on what their distros can rebuild or what they can build themselves. Not to mention they seemingly want to phase out the standard way of doing things in favor of the (in their opinion superior) GNU way of doing things, so everyone is more dependent on GNU tools and extensions and can't swap them out easily. Sure, they might think their reasons for creating this kind of vendor lock-in are more "noble" because they are doing it to enforce copyleft and open source on people whether they like it or not, but... well, there it is.

My earliest experience with Linux gaming was buying a copy of SimCity 3000 from Loki Games. It was very hard to use because glibc dropped support support for a.out right after it came out, and it required patches that were hard to find just to run on distros in the day. I did manage to get it working eventually, but that game has progressively gotten harder and harder to support on modern Linux to the point that most people will straight-up tell you that you're better off using Wine and running the Win32 version, because it is actually less complicated than getting the Linux version to work on a modern distro. So my experience has been that my Windows CD of the game has held its value, and the Linux one? Became basically unusable trash within 3 to 5 years of release.

[Tharthan]

Off-topic:

I did manage to get it working eventually, but that game has progressively gotten harder and harder to support on modern Linux to the point that most people will straight-up tell you that you're better off using Wine and running the Win32 version

I think that Mortimer Green wishes to have a few words with you about such wasteful expenditure, Mayor Athenian200.

[Moonchild]

My real question is.... Why?
What was so terribly wrong with the previous hash function that they had to write their own self-important "GNU" version of it instead?
It's a hash function, right? Hash functions aren't complicated and they are expected to do one specific and simple thing. How it's implemented is largely irrelevant.

Off-topic:
As a tangent, I think this is exactly what the main problem is with Linux overall - instead of working together to improve something, everyone wants to do it their own way individually. Hence there being so many different distros, 95% of them failing after a few years.

[jobbautista9]

DT_GNU_HASH was written because DT_HASH is slower and isn't good at finding missing symbols. This blog post seems to show how the GNU extension is objectively better than the standardized one.

I don't really mind GNU writing whatever extension they come up with; they should at least make sure there's good documentation behind it, and still maintain compatibility with the standards.

[Moonchild]

Hmm so it's about adding extra fields to a lookup table to make it faster.
That really doesn't explain why they just replace it without extensive periods of deprecation because it's effectively a different API. Even if it's just about performance they should provide a slower path for compatibility (in the deprecation period that doesn't exist)
So, IIUC it's a good idea to improve a long-standing function but was terribly mismanaged.

[athenian200]

The other concerns that come up are... the standard they replaced was part of a series of carefully defined Unix standards, POSIX, ELF, SysV ABI, etc. Instead of trying to make a formal proposal to change those standards, they just have glibc do its own thing and don't document it well, breaking GNU/Linux further away from the pack and making it that much harder to make stuff that works in a cross-platform way.

There's also evidence that other OS and libc vendors were definitely interested in DT_GNU_HASH, but were terrified to look at the source code and essentially had to reverse engineer it and rely on what little specification they put out to avoid potentially being legally bound by the GPL. So other operating systems have wound up either staying with DT_HASH, or creating their own extensions that are incompatible with each other. But as it stands, DT_HASH is still the only one that works on everything, and all the other stuff is incompatible and platform-specific.

I mean, I know it's not good to assume malicious intent, but it looks pretty bad that there's evidence of another OS wanting to implement something like DT_GNU_HASH and being careful to note that they "didn't look at the glibc source code" while the glibc team also doesn't write better documentation and makes it nearly impossible to understand their standard without looking at their code (which opens up legal liability for anyone who wants to write non-GPL code). And then on top of it they try to sneak in a commit that kills DT_HASH on Linux and hope no one notices? I don't know... hard to dismiss all of that as all being well-intentioned negligence.

[Moonchild]

Something can't be a standard if it's not or poorly documented. because that is what a standard is: documentation.

[mr tribute]

If you don't want to touch Windows or macOS then Linux is the only alternative on the desktop (banking and many other sites basically require Chromium). I see that Chinese distros like Linux Deepin and Ubuntu Kylin use Qt. They are quite "locked down", but really polished. I think Linux on the desktop will continue to fracture, but it will continue to be the only practical alternative to Windows and macOS.

If Chromium makes a Qt UI (there is some work going on) it might be the start of a slow death of gtk.

I think there is plenty of pressure on Google from Red Hat/IBM to not land the Qt UI pieces. This could be a game changer. Some Linux users might not need to have gtk installed anymore. Red Hat/IBM has a firm grip over Linux, but I think the Chinese can shake things up. They went from copying to basically become world leaders. Check out Chinese high speed passenger trains on YouTube for example.

https://chromium-review.googlesource.co ... /+/3583242

https://docs.google.com/document/d/1IO6 ... GX7JXGt-3Q

[Moonchild]

Off-topic:
Not to veer off-topic again but China is in a lot of economic trouble. Their prestigious projects were built on massive debt. Check out recent china news updates (on youtube as well)

[athenian200]

If you don't want to touch Windows or macOS then Linux is the only alternative on the desktop.

Actually, aside from driver support, I would say the alternatives really aren't much worse than an average desktop Linux. Other Unix-like systems (including my preferred alternative) have Wine, LibreOffice, etc. Sure, you can say that they are more like "Server OS" systems than as desktops, but that's true of Linux itself to a large degree. It's not as if Linux has made great strides and an alternative couldn't catch up to where it is quickly.

I get the kind of pragmatic, present-focused mentality your point is rooted in. But I think that if you are going to talk about alternatives to Mac and Windows, then it is only natural that some people will not be happy with Linux either. Besides... it's in many ways this deep-seated idea that no alternative besides Linux can thrive that keeps everything else back. Most people don't remember, but Linux was in the same position as most of the alternatives are now not that long ago, before Red Hat and then Google made big investments and basically gave it a shot in the arm with Android.

[mr tribute]

Off-topic:

Off-topic:
Not to veer off-topic again but China is in a lot of economic trouble. Their prestigious projects were built on massive debt. Check out recent china news updates (on youtube as well)

Regardless, China needs Windows replacements and they can use Qt for free (in FOSS projects) as a building block so I see a continued demand for Qt in China. Will Chinese consumers switch from Windows? Not likely, but just government and local administration is huge and represents millions of computers. The next release of Linux Deepin will be a "full OS" with only a few select pieces from Debian. It will be Debian compatible because that is a big advantage for third party developers. Apart from that, Linux Deepin doesn't really need Debian anymore. They also created a next-gen package format that can be used alongside deb packages. They push ahead. My only concern is how open/transparent this OS will be. Dependencies will probably also be an issue (just like KDE), if you want something lightweight that doesn't install 500 packages.

[mr tribute]

If you don't want to touch Windows or macOS then Linux is the only alternative on the desktop.

Actually, aside from driver support, I would say the alternatives really aren't much worse than an average desktop Linux. Other Unix-like systems (including my preferred alternative) have Wine, LibreOffice, etc. Sure, you can say that they are more like "Server OS" systems than as desktops, but that's true of Linux itself to a large degree. It's not as if Linux has made great strides and an alternative couldn't catch up to where it is quickly.

I get the kind of pragmatic, present-focused mentality your point is rooted in. But I think that if you are going to talk about alternatives to Mac and Windows, then it is only natural that some people will not be happy with Linux either. Besides... it's in many ways this deep-seated idea that no alternative besides Linux can thrive that keeps everything else back. Most people don't remember, but Linux was in the same position as most of the alternatives are now not that long ago, before Red Hat and then Google made big investments and basically gave it a shot in the arm with Android.

I think we are having two slightly different discussions at the same time. I do think I understand what you are saying. What I want to say is this: It doesn't matter if Sun OS is the world's best OS. If it doesn't have Chromium it is a no-go for practical purposes. There is a reason you use Windows, isn't there?

[andyprough]

I don't think this really affects Pale Moon

I'll assume that this is the key takeaway for someone like me, since I couldn't tell the difference between a DT_HASH table and a DT_GNU_HASH table if my life depended on it. If it ever becomes a problem for Pale Moon, please let us know.

[andyprough]

It doesn't matter if Sun OS is the world's best OS. If it doesn't have Chromium it is a no-go for practical purposes.

Not having chromium and electron makes it more attractive in a way. I'm assuming Sun OS is long dead - isn't it OpenIndiana or some modern fork like that?

[mr tribute]

Off-topic:

It doesn't matter if Sun OS is the world's best OS. If it doesn't have Chromium it is a no-go for practical purposes.

Not having chromium and electron makes it more attractive in a way.

Again, two different discussions:

1. Ideal world
vs
2. Practical reality

I think we all agree that it is wrong that Google has a monopoly on the Internet. On the other hand, if you want to get something done on the Internet you cannot shield yourself from reality. In the end reality wins and we all have to bite the sour apple - start Chromium.

What I tried to illustrate in my posts is that the situation is similar with gtk. Gtk has some sort of monopoly in the desktop Linux space. It doesn't have to be this way. In fact, this monopoly is much easier to undo than the monopoly Google has on the Web.

I know that all this is slightly off topic, but I'm tired of "Gnome broke this, Gnome broke that". Red Hat created desktop Linux. Now they destroy it. It can be seen as "natural evolution" in a fallen world and Microsoft is basically treating Windows the same way, only a little bit more careful since they have paying customers.

As far as Pale Moon is concerned it would be interesting to know if someone is using Pale Moon on musl:
https://musl.libc.org/

[andyprough]

Again, two different discussions:

1. Ideal world
vs
2. Practical reality

Agree to disagree on gnome and chromium - I've never used gnome other than to try it momentarily and I don't need a chromium-based browser for anything I do online. But that's OK, it's good that we have different experiences, and I agree that what you describe is probably the reality for a sizable portion of desktop Linux users.

As far as Pale Moon is concerned it would be interesting to know if someone is using Pale Moon on musl:
https://musl.libc.org/

I tried it on a Void Linux musl vm and Pale Moon wouldn't start up for me without creating a glibc chroot for it. So basically no, not without figuring out how to specifically build it for musl.

[mr tribute]

Agree to disagree on gnome and chromium - I've never used gnome other than to try it momentarily and I don't need a chromium-based browser for anything I do online. But that's OK, it's good that we have different experiences, and I agree that what you describe is probably the reality for a sizable portion of desktop Linux users.

I don't know why you mentioned Gnome because I have never tried Gnome(3) either. Maybe Firefox can replace Chromium in some cases. Personally, I never use Firefox because I see it as just another Chromium clone (even if this is technically incorrect).

I tried it on a Void Linux musl vm and Pale Moon wouldn't start up for me without creating a glibc chroot for it. So basically no, not without figuring out how to specifically build it for musl.

I see. glibc is required for gtk I guess. I checked Debian dependencies for LXQt. I can't see any obvious dependencies on gtk or glibc (distro base system will likely pull it in though). Regarding user-facing software it is basically browsers that force Linux users to use glibc/gtk I think.

Pale Moon is good with gtk3, but a gtk4 port isn't possible without adding "external" (self developed): menus, menu bar, status bar etc. Gtk4 is a full transition to a "convergent" paradigm that supports touch, keyboard and mouse (probably in that order of importance). Menu bars are non-existent. I don't know if traditional server-side title bars are possible in gtk4. Menus exist as "popover menus", but traditional menus are gone.

While glibc changes might be of more immediate concern what I am trying to say is that there is no future for traditional desktop applications using the gtk "toolkit". It can be done with lots of custom code like the gtk4 port of LibreOffice, but what is the point of using a "toolkit" if you have to write half of it yourself?

[athenian200]

If it doesn't have Chromium it is a no-go for practical purposes. There is a reason you use Windows, isn't there?

Well, actually I would say I use Windows more for gaming and Visual Studio than for Chromium. There are probably more sites I use that still require some variant of IE (or at least an ActiveX control) than require Chromium. I also wind up needing Microsoft Office a lot, plus I have a lot invested in Windows Store apps over the years, etc... Also, it's not as impossible to get Chromium ported over as you would think (someone did it once, the port just wasn't maintained)... it's just no one has bothered because Firefox is extremely well-established on here (they actually put months of work into porting Rust of all things and succeeded), WebKit is also available and it's "good enough" (Safari on iPhone still uses that), and no one willing to go to the trouble of using an obscure alternate OS actually wants to do that just put a lot of work into supporting Google Chrome.

I'm pretty sure Chromium could be ported, though. We have all the dependencies, it's just no one using the OS actually wants it... hard to believe, I know. To be fair, I've known plenty of Linux users who refuse to install Chrome or Chromium-browsers as well and insist on sticking with Firefox. Not that I think that is particularly productive or meaningful, just that it's more common a mentality than you might think. There's a reason why many Linux distros still seem to ship Firefox as the default browser.

Not having chromium and electron makes it more attractive in a way. I'm assuming Sun OS is long dead - isn't it OpenIndiana or some modern fork like that?

Two modern forks of Sun Solaris... Oracle Solaris and OpenIndiana. I tend to call it SunOS because that's what the kernel is called (SunOS 5.x specifically), and Solaris refers exclusively to Oracle's commercial fork. Plus it gets tiring to type Solaris or illumos-based operating systems all the time.

I see. glibc is required for gtk I guess.

It's not. OpenIndiana and Solaris have GTK available, and they do not use glibc.