This is a big forwards-compatibility risk. Suppose glibc adds a new symbol, and then a GPU driver adds a dependency on that symbol. The user wants to run an old executable with the updated GPU driver (maybe the old GPU driver doesn’t support their GPU). Normally, this would work fine: the user has to use a new copy of glibc, which will be compatible with both the new GPU driver and the old executable. But with your approach, the GPU driver is forced to use the glibc reimplementation which has been statically linked into the executable. Which, since the executable is old, can’t possibly implement the new symbol.
The same issue would occur if glibc adds a new version of an existing symbol and then the GPU driver is recompiled. (Or, for that matter, if a GPU driver adds a dependency on a symbol which glibc has always supported but which isn’t in the subset that you reimplemented, though in theory that could be solved if you reimplemented 100% of the symbols.)
Yes this is just asking for trouble - and all it does is solve a problem that doesn't really exist. Just dynamically link against the oldest glibc you want to support. Its annoying that Linux toolchains don't have built in easy mode support for that but its much easier to deal with than this thing will be when it breaks.
It's also not just new symbols, the loader semantics also aren't static and new enough libraries may not support older semantics - e.g. the loader used to use DT_HASH entries for symbol resolution but now they are no longer present on all distributions.
This is not a desirable solution on musl based systems. Whatever you do in that situation ends up horrible, so it is about finding the least bad solution. Which this seems like a workable variant of.
> Just dynamically link against the oldest glibc you want to support.
Just the other day I tried running an older binary and it failed with a glibc error, despite it being linked to a glibc version that's barely 5 releases behind the one on my system. So maybe glibc isn't backwards compatible after all...
So we are re-inventing patched a.out files, back when UNIX systems started to introduce dynamic loading, before ELF was invented?
Advocates of static linking keep forgetting once upon a time UNIX only had static linking, then we had overlays, and eventually dynamic linking came to be.
And I also remember very well that dynamic linking appeared ONLY because we were catastrophically short on memory; everything else was added much later.
Now we have plenty of memory, and we can very well return to our blessed roots!
I have faced a similar issue in the past, and I don't understand how static binaries from the host are supposed to solve this.
From what I remember, GPU access on Linux 'works' by accessing specific FDs under /dev, which are vendor specific - this is what these libs do under the hood.
The libraries don't have any magic powers - if the FD is inaccessible, you won't be able to do anything.
So there's some vendor specific access needed in containers anyway (or a blanket allow, which is a BAD idea).
Also not sure why dynamic linking isn't good enough for this - the issue lies with the permissions, not how you load/link libraries.
If you can figure out your own ELF loader, you can figure out how to build a partially static executable that doesn't need this. You can mix static and dynamic linking. Build tooling around that is just shit.
In this scenario, I'll have to choose which libc I want to run. These won't be portable Linux binaries in the true sense of the word; I'll have to leave Alpine out, and possibly Android, which I don't want.
Can you really mix it? When I break or remove a shared lib, some binaries no longer work. With static libs or even better, e. g. statically compiled busybox, I don't have that issue, so I disagree on the claim that mixing solves everything as such. I keep the basic toolchain I use as statically compiled variant. The whole system works better if I can break it less easily.
I've implemented the same thing for micron (more or less). One advice I'd give you is to _really_ take care regarding SysV/ELF ABI conventions, there's tons of undocumented stuff in there and it's really easy to mess something up or cause a security defect (see AT_SECURE). That being said the way you're doing is also tricky(ish) because if I understood your implementation correctly you're hooking this into an already running musl which could cause backwards compatibility issues if musl changes under you. Doing this is safer if you control the entire runtime.
> One advice I'd give you is to _really_ take care regarding SysV/ELF ABI conventions, there's tons of undocumented stuff in there and it's really easy to mess something up or cause a security defect (see AT_SECURE)
Yes, it's not trivial, but I hope that over time everything will settle down.
> if I understood your implementation correctly you're hooking this into an already running musl which could cause backwards compatibility issues if musl changes under you
No, I hook this to musl, which is statically linked into my binary, and I have complete control over it.
Every couple of years, I revisit my PL dev hobby and this time I decided to create a language/runtime with pre-emptive scheduling using instruction fuel. While I always do freestanding builds, this time I decided that I also wanted to support native FFI.
That is when I realized the true horror of (g)libc. It wants to inject itself at the root of the library/program and everything from threading to dlopen/dlsym is impacted. I tried a lot of workarounds including trying to implement a loader myself, but the complexity (and fragility) grew so much that I felt it was not worth it.
Finally, I retreated into the safe world of a freestanding runtime + syscalls. FFI, if it has to happen, will occur via IPC of some kind. A second process linked against glibc that will manage calls on behalf of the clean first one.
> GPU: Vulkan and OpenGL drivers are supplied by the host as shared objects, usually built against glibc, and a fully static musl binary cannot normally dlopen() them.
Why? Have people managed to break the ancient concept of shared libraries, and this is a fix for that?
Because glibc and GNU set a terrible precedent. On GNU/Linux systems the shared binary interpreter / loader, GCC compiler, the C library and the system C/C++ ABI all depend into each other. You cannot change any of them independently. All shared libraries depend on the specific glibc version to load them into memory to be able to use that specific glibc version as their C library and make calls like dlopen.
Shared libraries have always been broken in Linux. Unfortunately many things like GPU drivers, graphics libraries and NSS need shared libraries to dynamically load certain runtimes (because you don't want to load all possible GPU drivers in existence to your RAM). So an ecosystem has been developed on top of terrible ABI and architecture GNU/glibc provided.
Yeah, it's mind boggling how everything hard depends on glibc, even critical graphics systems.
I've become obsessed with getting rid of it, especially after I realized that contributing to GNU itself was a dead end. Freestanding Linux programming turned out to be much more fun anyway.
All libraries out there should adopt the SQLite design: programmers provide it with all the necessary functions. Instead of libraries hard depending on glibc, we get to inject the libc-ish subset it needs. Then we can use whatever we want under the hood. I'm working on porting SQLite to freestanding Linux system calls so it can run with zero dependencies. Wish I could say the same for software like mesa, I'd need a lot of help for this one...
Modularizing (and versioning each part independently) libc would go a long way. There is also a need to separate the stuff needed for system integration with what is necessary for users of the C language to actually do stuff.
Can't be done. The libc is legacy, it can't be changed without breaking everything. It's also mandatory on every operating system other than Linux.
A change in paradigm is necessary. Freestanding C, not hosted C. This completely gets rid of the libc and is a surprisingly clean language. Linux only, because it's the only kernel with a stable binary interface. Every other OS forces a C runtime.
I once worked on a liblinux project that embodied this... Stopped because Linux itself has a nolibc thing in the kernel tree and I didn't want to compete with it. Now I'm working on the Rust version.
> what is necessary for users of the C language to actually do stuff
Surprisingly little. I wrote an entire lisp interpreter in freestanding C with Linux system calls. It managed to survive for a rather long time without any memory allocation at all.
The system layer is refreshingly tiny. It consists of a memory allocator and extremely basic functions like memmove and strlen. I successfully got rid of total nonsense like thread local errno, locales, implicit buffering, cached global state, possibly more. All that stuff is gone! Exactly one global survived: the stack canary generated by GCC and clang. Every other symbol in the ELF is controlled by me.
Wasn't able to get rid of the NUL terminator. Linux itself needs it. To get rid of that little billion dollar mistake requires an entirely new kernel with zero UNIX/POSIX influence. I had to make my peace with that one. All my buffers maintain an extra NUL byte at the end.
>A change in paradigm is necessary. Freestanding C, not hosted C. This completely gets rid of the libc and is a surprisingly clean language. Linux only, because it's the only kernel with a stable binary interface. Every other OS forces a C runtime.
I'm sure those OSes make efforts to make said runtime binary compatible between executables.
> A change in paradigm is necessary. Freestanding C, not hosted C. This completely gets rid of the libc and is a surprisingly clean language. Linux only, because it's the only kernel with a stable binary interface. Every other OS forces a C runtime.
Great choice for small programs, but what if I want hardware accelerated 3d?
Yeah, that's the annoying part. Been wondering about this for years, and graphics support was among the first issues raised on the lone lisp GitHub repository. At this point I've even started exploring the mesa codebase, made some patches but didn't submit them yet due to the AI stigma.
With Linux system calls alone it should be possible to set up kernel mode setting without depending on any toolkit at all. This should be enough to get a framebuffer for software rendering.
For hardware acceleration though, one must give this graphics context to an OpenGL ES implementation. That's where it gets ugly. There is no way to divorce that from the libc short of literally rewriting it.
Maybe Vulkan will enable it? I can't say for sure at my current knowledge level.
> Can't be done. The libc is legacy, it can't be changed without breaking everything. It's also mandatory on every operating system other than Linux.
Windows explicitly does not want you to link the system libc. You are expected to bring your own, and doing so means your process has multiple libc's loaded into its address space.
And if you choose to build a binary that doesn't need a libc, you won't be bringing one.
The Windows ecosystem, that manages to deliver built binaries easily & widely, regardless of whether the author has a 1 year old OS or a 15 year old OS, suggests that it's not as big a problem as you believe.
It's not a problem in the same way that things like snap or flatpak aren't problems. It works but it bloats things up considerably and makes you wonder where it all went so wrong. I mean, dozens of slightly incompatible runtimes inside a single process?
Btw, absolutely insane that it's 2026, and Linux cannot do the most defining OS thing - that is, provide a standardized environment to run binaries against.
All solutions to this problem are hacky, complex and controversal and highly fragmented, where this should be BASIC functionality
Wasn't Linux Standard Base supposed to fix this? I cannot imagine any reason why glibc would break at a rate that you can't keep the current version binary compatible for years.
Or do the MS thing, and ship multiple versions like msvcrt
That's libstdc++ depending on a C ABI (or a GNU ABI to be precise) but that doesn't make it a C++ ABI. You can also have global constructors in C with GCC.
Somehow most of my portability issues seem to be caused by glibc, its symbol versioning and close ties to the dynamic loader. Minor versions aren't compatible, no two Linux distros ship the same version and you can't just provide your own without also patching in your own dynamic loader.
At least as far as the defaults on Linux go I consider C the root of all evil.
AFAIK glibc is backwards compatible as long as a) your program is running on a newer version (i.e. you're not trying to dynamically link against an older version) and b) you're not using hidden/undocumented symbols.
In which case as long as you're using the documented public API and compile your program with the oldest version of glibc you want to support (some Ubuntu from 4-5 years ago should cover pretty much every current desktop) you should be fine. And with something like Docker this is trivial to do.
Sure it is annoying that you cannot use your current distro (especially if you use some rolling distro) to make binaries for everyone, but it takes very little effort to work around that. The only issue i can think of is if you absolutely want to compile using the latest version of your compiler and you cannot build the compiler from source to work in the Docker (or whatever) contain to work against the older glibc.
In complex cases, it turns out that the old version of glibc also pulls in other libraries and the compiler, and you're stuck with a very ancient sysroot.
You may often find that you can't compile new library versions in such a sysroot and link them statically.
So, it looks good on paper, but forget about the ravines.
While I don't disagree with some of the pain you describe, you conveniently gloss over the fact that gnu developed a system that worked, and then made it free to everyone to consult and use.
BSD also did it. They did it better. Maybe more modern but AOSP also did it but at a different level of binary: instead of ELF, using compiled Java bytecode archives.
The root cause is alternative libc implementations. Are BSD syscalls considered stable? I remember Go moving to use libc on OpenBSD. Solaris also has the libc as the stable interface. Linux kernel is an outlier here guaranteeing stable syscalls but you wanting to use another libc is not Glibc's problem.
Except the other UNIX systems, starting with the original one, evolved from only having static linking to various ways to connect libraries and applications.
In what sense do binary interpreter / loader, GCC compiler, C library and system C/C++ ABI dependent on each other? I have certainly mixed different versions of all these components without problems so far.
When you compile GCC you need to provide a full glibc installation as your target. It is also a dependency of libstdc++.
C++ global/static variable initialization depends on the specific version of glibc (they don't usually break compat, but they can and they did in the past) which also provides ld-linux.so that loads those global variable placeholders in the correct manner such that glibc and libstdc++ can initialize them correctly.
This is just one example. Thread local variables and behavior of things like pthreads with signal, fork etc all depend on glibc.
I can't comment on the C++, I can imagine there plenty of issues, but for C I don't see this. You need some libc if you compile with gcc, but this generally does not introduce a hard version dependency on the specific version (there may be a minimum requirement if you compile against a new version that a symbol with a different ABI).
I don’t imagine that you’re unaware of any of this, but: ld.so and libc.so are heavily interdependent in deliberately undocumented ways with Glibc and outright the same file with shared Musl. And while you might usually get away with using any old GCC with the right architecture and ABI (especially for C; cf the musl-gcc hack), technically it needs to be built to target a specific libc version (particularly via symbol versioning; I’ve long wanted to gather a set of patches to build an old Glibc and subsequently a cross-compiler using a new GCC so I could avoid PyPA’s manylinux monster or its moral equivalent for compatible dynamic binaries in simple cases). The C compiler of course is tied to the C ABI, and this wouldn’t be really worth mentioning except for the time where the GCC devs accidentally the whole SysV i386 ABI and pretended that the stack was always 16-byte aligned, why do you ask, except on RHEL. The C++ parts I can’t really comment on.
I am not really sure. For ld.so and libc.so I may believe this. The C ABI is very stable, and if you use a new symbol from a newer glibc, you certainly depend on it, but this can also be avoided. In any case, I do not see what is fundamentally misdesigned here. I can't quite image how it could work differently. If you upgrade something so that the e ABI changed you natually need to update other components. Static linking certainly seems a very poor replacement for this.
> I’ve long wanted to gather a set of patches to build an old Glibc and subsequently a cross-compiler using a new GCC so I could avoid PyPA’s manylinux monster or its moral equivalent for compatible dynamic binaries in simple cases
And that's the correct approach and also one that many have taken. We just need someone willing to maintain that as an easy mode SDK for everyone.
Who said that? This approach has many problems that have already been discussed here, not to mention the fact that it leaves Alpine and Bionic-based systems out in the cold.
Let me remind you that Bionic is the most widespread libc in the Linux world, and Alpine is the most popular Docker layer.
The world doesn't end with glibc. And it doesn't begin with it.
The interpreter/loader is glibc and a key part of bootstrapping an executable built against glibc is loading libc itself before continuing on to load the program. Versioning is a problem when distributing binaries linked against a newer glibc to distros that ship an older one. The C compiler doesn't really care as much.
Until you define a thread local variable (C11) or use atomics (also C11) or define a global with an initial value. Then it happily generates code that depends on "whatever my target glibc + ld-linux.so needs".
You'd expect that but, no. That's why you cannot load glibc-linked binaries in a Musl distro. Edit: that's why the hacks like the original post is needed, as well.
The ABI is strongly dependent on explicit libc implementation in current Linux systems. There is no libc independent ABI on Linux.
Sorry, can you be more specific. I do not understand what the problem is. If Musl does not implement support for the ABI, this would be a musl problem?
There is no libc independent ABI. ABI doesn't purely mean just calling conventions.
When you compile libc, you also get a binary loader ld-linux.so with it. They are not two independent components of a system.
Basically all .so files compiled with glibc require the ld-linux.so that's also generated by that glibc (or a later version, if they didn't break the binary compatibility).
There are a lot of stuff that's executed by ld-linux.so and glibc that are not explicitly documented but they are absolutely necessary for your program to start and correctly initialize things like global variables or signal handling or loading other dynamic libraries. Some of that functionality sits in ld-linux.so and some of that in glibc. They have circular dependencies to each other. glibc expects ld-linux.so to put things in certain order but ld-linux.so also must load glibc first to have access to certain APIs. They are not part of System V ABI. They are not documented.
Musl maybe can implement this but it is simply reverse engineering what glibc did and then playing a game of cat and mouse. There is no independent ABI standard.
Sorry, again this too vague for me. What is the exact problem with atomics and thread_local in C that would make the ABI dependent on a specific version of glibc? I know the ABI is not just calling convention and e.g. for atomics may involve calling a function from libatomic. But from my understanding, this is all part of a standardized ABI that does not change and can be provided by different implementations.
Then, what is the exact reason a library compiled against glibc must be loaded by a specific ld-linux? I could see that this is true for C++ perhaps, or when you use very special features, but I do not see this for C.
I often compiled programs against one version of glibc and run it against a different version, so I know there is not a tight coupling. So please be specific in explaining in what scenarios this would break.
Most everyone else is talking about the problem while you mapped the room.
Solo is basically pg83's answer to the architecture you just described. If there is no libc independent ABI, build your own loader and shim the boundary.
I can understand Linus's obsession with taste and the areas it was overlooked or traded.
> all shared libraries depend on the specific glibc version to load them
Not really, though. glibc uses symbol versions that are forward but not backward compatible. If you got an error that said "this program was built for a newer version of <distro>" would you say the same thing?
Note this is the same (if not worse) on MacOS, and on windows you used to distribute the CRT with your application just to deal with the same problem.
Yes glibc has some backwards compat but you cannot load a binary compiled with a newer version of glibc using an older ld-linux.so. That's because the interdependency. Nor you can load binaries that depend on different libc.so files with glibc systems
I cannot comment on macOS, I have never used it. However this is not a problem with Windows. You can ship a newer CRT or you can install it as a system component using Microsoft's MSI. The dependency is one way on Windows. CRT purely depends on Win32. Moreover the loader is completely independent and DLLs are loaded into their own unique scoped namespace unlike Linux that loads them in global symbol namespace. That's why you can mix and match DLLs compiled for different CRT versions.
It is not just compatibility. You cannot load them into the memory with your system dynamic loader. You need to also ship ld-linux.so with the new version of glibc you have, if you were to distribute your program independently.
On Windows you don't need to ship a new binary loader. I can just ship Windows 10 UCRT DLL (which is the new libc of Windows) to Vista and my binaries will work. The binary loader isn't interlinked with the libc.
Windows loader is part of the system ABI, and programs' libc(s) are loaded by it.
One of the ways Windows manages to support multiple libc's is by being careful not to mix allocators; if a system API you call allocates on your behalf, your libc can't free it, the system API will offer a function to free it.
Windows loader is certainly available to user programs though; LoadLibrary has been around longer than many developers.
LoadLibrary will also not be able to load arbitrary libraries compiled for newer versions of Windows though. Just like with gcc, Windows also does not guarantee forwards compatibility - because that would men freezing the feature set the system libraries provide.
Windows handles this much better than any other OS. The API passes versions (== structure sizes on the calling side), and WinAPI can handle that.
As for older Windows systems not being able to load new DLLs, they can; the format hasn't changed in a very long time. I've had experience with installing some DLLs on Windows NT 3.51 and running a modern Firefox, which is about 20 years behind the times.
Targeting an older version of Windows is "set a define so that the system headers don't expose functions that didn't exist on older Windows". You don't need an old toolchain to target old systems, you ask the newer toolchains to target it.
MS haven't made this work arbitrarily far back, I believe they deprecated targeting Windows XP in one of the more recent toolchains, but that was purely a "not worth supporting" situation.
In what way are they "broken" when Linux runs fine on millions of boxes? Sure, it might be a pain for proprietary software, but if your app is open source it's not that hard to build it on whichever distro you want. If your app is popular enough the distro maintainer will build it for you
Because many folks don't understand UNIX systems introduced dynamic linking for several reasons, and they actually only had static linking for almost 20 years, since UNIX was known outside Bell Labs.
Additionally many other OSes have had both approaches since their early days, Xerox PARC ones.
For some strange reason they assume to know better than all those researchers.
These decisions, and these studies, were made a VERY long time ago. It's completely unclear why the decisions made then are relevant now, and why they can't be challenged.
It is like advocating that we should drop cars and go back to chariots, because wooden wheels don't get flat, while forgetting why they are mostly used for tourists nowadays.
Well, now it's possible! Furthermore, SoLo binaries can run, without modification, on glibc-based distros, alpine, and soon on android/bionic (not committed yet).
> Why? Have people managed to break the ancient concept of shared libraries
If you break or remove a shared lib here, you may no longer be able to compile something from source. I had that happen in the past before I started to use more statically compiled programs (and busybox too).
Assuming everything works as-is via shared libraries at all times, makes little sense for ALL linux systems. For instance, some people upgrade glibc manually. Then you need a working base system to resume compilation. I do that for my customized gobolinux system, so I can use any program version as well as any glibc version (assuming I can still compile the program; many older programs no longer compile).
On the one hand, this is technically true, but on the other, what serious issues do you know that will cause problems in practice? I run tests on 1,000 of the most popular Debian packages.
> Do you belive the machine is taking offence? It can not.
Obviously, this is an offense to me.
> Telling me, a coder you never met or interacted with before, that I would introduce more bugs than The Product(tm) is pretty rude and prejudiced.
Don't exaggerate. I didn't say that you personally introduce more bugs (as you rightly pointed out, I don't know you and don't know how often you introduce bugs), I said that the average developer introduces more bugs than an SOTA LLM.
I don’t think I’ve said it out loud more than a couple times in my life. But in general I think I spell out / pronounce the “dot” in file extensions unless it’s completely obvious from context.
Technically, you're right, it's a dynamic loader. Technically, it's pure dynamic loading.
If we look at the issue at its core, we're still a statically linked program in a hostile environment, forced to dynamically load device drivers from the system.
It's similar to Golang; on MacOS, it has to use libSystem, even though otherwise, these are the statically linked Go binaries we're used to and love.
Let me add a little more detail: if I use vdso with gettimeofday in a statically linked program on Linux, am I still a statically linked program, or not? :)
It is a testament to the complete failure of the GNU/Linux userland that something like this seems at all attractive to spend time on (or, it seems, LLM tokens).
Actually, scratch that, because Windows and macOS have historically struggled with ABI compatibility as well (macOS less so, due to not caring about backward compatibility in the first place).
How did we get to the point where people feel they need to go to the length of embedding an ELF loader in their binary (!!) rather than just linking with glibc?
> How did we get to the point where people feel they need to go to the length of embedding an ELF loader in their binary (!!) rather than just linking with glibc?
Most Linux distros have been built around the ability to compile their software together in a large repository, from source, so this was rarely ever an issue. Proprietary distribution or executing binaries from the internet like on Windows just wasn't really a common issue.
The problem arises when you start combining distros (glibc and MUSL for instance) or if you try to do the Windows model of sharing software. Historically, projects just compiled different versions for different distros.
When doing static compilation, just targetting an old version of glibc (which is generally forward compatible) also works.
You can hack your way into using software like this (or rather, have an LLM hack its way in) but I don't think any real distro actually cares. This issue exists in a quite small space where people are trying to use proprietary software built for glibc in MUSL environments for whatever reason, and the usual compatibility tricks don't work.
It's a niche use case for most Linux distros. It's not a "complete failure" of the GNU/Linux userland, it's the result of a couple of proprietary components not having MUSL builds available, or MUSL-based distributions not including libraries people want.
I'd like glibc to change so that these hacks aren't necessary for these use cases anymore, but it's not really a problem in practice for the vast majority of Linux use cases.
Yes, on the one hand, each specific distribution doesn't have this problem because it can pick up everything it needs.
But for us, independent developers of small programs, the problem is truly stark: we can't afford to build our programs for every distribution. And we can't afford to waste time navigating all the idiosyncrasies of various package repositories, both technical and political (not all repositories allow easy access).
And we can't count on someone else packaging our work until we become incredibly popular.
I'm mostly taken aback all the solutions devised to go around the issue, especially the container-based ones. I really disliked it when I grabbed the flatpak version of Blender only to find out that it can't have HIP support. (they might have fixed it by now but you get the point)
Linux loves to leave papercuts unfixed or undocumented for decades. The solution is to build against an older version of glibc but no one tells you that or how to do it.
Luckily Zig makes this quite easy to do. In Mach[0] we are able to just `zig build -Dtarget=x86_64-linux-gnu.2.28` to build GUI apps against an ~8 year old glibc version for maximum compatibility.
This is possible because Zig allows for targeting most glibc versions out of the box with its cross-compilation support.
1) Why should I limit myself to the available APIs?
2) Not just glibc. For example, if I build against the latest libstdc++, it will automatically support the more recent glibc. And pinning the old libstdc++ -well, that's just not a good idea.
Complete failure is strong words when lots and lots of Linux boxes are running just fine.
I think the disconnect is mostly people that can't decide if they want a stable distro or a rolling release distro. Most everyone uses a stable distro because it's stable, but then the want some up-to-date software that isn't ore-built for their (crusty old) stable distro and they get annoyed. My solution was to finally give in and embrace a rolling release distro (I use arch, btw). If there isn't a package for something I want, it's not hard to build something myself because all my build tools, kernel, and libs are up to date.
Other reasons to want static linking is to distribute proprietary software with no source code available. Linux certainly does not cater to that scenario and I suppose some might call that a complete failure ¯ \ _ ( ツ ) _ / ¯
Glibc has a terrible history of binary incompatibility. If that's so hard to believe, try running binaries built on one distribution on other distributions. Linux has two stable ABIs: the kernel ABI for static programs, and, ironically, WINE.
I haven’t heard of this and I don’t think you’re right. Glibc, for all its faults, as a general rule does backward compatibility well. The problem is if you compile against a newer glibc (common in CI by default) and try to run on a distro with an older (common in the wild). If your CI uses an older glibc you should be fine AFAIK.
Not sure what you’re trying to show with that bug report but it’s not a case of cross distro glibc issues. If I read correctly it’s a vanilla behavioral change that exposed preexisting UB in flash.
Not sure how the comments about alpine or bionic relate either to my claim that cross distro glibc is fine.
It depends on how we define the ABI. I see it as a set of client-visible invariants that they rely on. In my world, glibc changed the client's visible invariants, breaking the client. The client works on one glibc-based host, but not on another. What is this if not "a case of cross-distro glibc issues?"
Overall, both of our points of view on compatibility were discussed well in that thread; we probably shouldn't repeat ourselves. :)
ABI is not "whatever happens to work with this distro" but "what programs that comply with the API contract compile to". Overlapping memcpy arguments is an API contract violation and thus not something covered by the ABI either. This distinction is the entire reason why C has a separate memmove function. You can't just make up your own imaginary ABI contract and then blame the system when it doesn't fulfill it. That's going to result in self-inflicted plain on any OS.
This is not true. Glibc supports symbol versioning. You can use it to select old versions of used symbols. The result is a binary that can work on 20 year old distros the same as on the latest, compiled with latest compiler and Glibc.
You can also compile using old distro and old Glibc to get similar effect. Though you would miss the advances of the newer compilers.
> How did we get to the point where people feel they need to go to the length of embedding an ELF loader in their binary (!!) rather than just linking with glibc?
Due to FUD, mostly.
Sane people do just link with an old enough version of glibc.
How are we supposed to take this stuff seriously if the author (sic) isn't even willing to write the readme? Claude exists! If I want some slop I can push the button myself.
Disclaimer: This is a user’s perspective rather than a programmer’s perspective.
valid point. I am usually okay with LLM generated code since even if it might not be architecturally sound It is usually well commented and has tests and documentation for helping another agent/human debug any issues.
But, just the painful experience of debugging any dlopen related crashes and/or intermittent bugs; and the sheer amount of tokens burnt by an LLM chasing tangents when shown a stack trace; I wouldn’t touch this at least as a packager/consumer of certain apps for personal usage on older distros.
So far, AnyLinux-Appimages seem to be a mature solution with great support from the developers, in case anyone lands here for packaging applications to run on older distros.
Modern models, when properly managed with a human in the loop, write higher-quality code than humans and introduce significantly fewer bugs.
Therefore, it's quite the opposite - you should expect fewer "dlopen-related crashes and/or intermittent bugs."
> So far, AnyLinux-Appimages seem to be a mature solution with great support from the developers, in case anyone lands here for packaging applications to run on older distros.
The Linux ecosystem already settled on containers to solve the problem. Namely Docker, Flatpak, Podman ec.
README.md was written by me, and I, of course, used claude/codex for it. In general, I do everything through claude/codex, the reasons are described in https://github.com/pg83/solo/blob/main/CONTRIBUTING.md . And no, it's not low effort, and no, I don't see the point in wasting time de-claude-ifying the text just to avoid it looking like I didn't spend enough time on it.
You make it clear why you write all your code through a llm. But a README is not code. Presumably you would like people to read it. A machine authored readme reflects poorly on a project.
In any case, it's open source. If you don't like something, even if the project seems generally useful, go ahead and fix it. The PR came in. I'm an engineer and I can write good code, but that doesn't mean I can write good README.mds!
The author of the README is me, the machine just wrote it.
I am not a native speaker of English, my written English is simply terrible, no one wants to read the text that I wrote exactly :))
For context, I read Claude output every single day, I know what it is reliable with and what it is not. Or at least I have a feel for how much I can trust it.
It may not be clear to a non-english first language person, but when I read Claudes documentation, my brain immediately picks up claude-speak. Therefore, I expect the code to be generally correct, maybe, depending on how specific I was during my prompting. In no way shape or form do I really trust it, at least until i dig into the code and validate my mental model. And query the review for edge cases etc...
By writing your documentation with Claude, my brain immediately associates the quality of the project with the quality of unreviewed Claude output.
To a degree this is unfair, as it is like judging the quality of someone's work based on their accent.
But Claudes accent, has a high correlation with Claude, so unlike with people, where an accent has no bearing on technical ability, Claude being Claude does.
I would much rather read a typo ridden sentence than claudism, even just as a forward, explaining what you did vs the ai, and telling the users how much we should trust it.
Also, If you really insist on using AI to write, have another model rewrite docs/comments into regular English (opus 4.6 for example is much better than 4.7, 4.8, or 5.
It is open source so you can do whatever you like, but people (especially native English speakers) will discount your work, because Claude, especially opus 5 writes very very badly.
The people who say they would rather see a typo infested mess of a readme are serious. Or even just write in your native language and then have Claude translate.
Both of those are better indications of proof of effort than a Claude readme.
> By writing your documentation with Claude, my brain immediately associates the quality of the project with the quality of unreviewed Claude output.
I'm sorry, but I agree with the author: if a certain writing style makes you associate the work with low-quality, then that's your problem. The author shouldn't have to rewrite the readme just to avoid triggering your automatic unfounded associations. If you look at the substance of the work, including the test cases, then this is clearly not easy work that can be vibe coded in a single pass.
It's just like emdash. Everybody digs on how it's a signifier of LLM text, but I've used emdash for years because it's gramatically correct. I shouldn't have to stop using emdash just to avoid kneejerk reactions.
LLMs often use emdashes in a distinct incorrect way. It's not just the existence of any emdash, although considering you were in a very small minority of older users, it now warrants increased scrutiny, unfortunately for you.
if 99 poor effort/quality projects have a readme that reads in a particular style, then you expect the 100th project with a readme in the same style to also be of poor effort/quality
statistically, it only makes sense for your expectations to immediately be low when you encounter this writing style because there are just so much slop out there
the author is free to keep that writing style but they should be aware that this will—at least on the surface level—make their project look exactly like the metric ton of slop we see posted everyday everywhere
I believe usually when someone complains about text written by a language model they are hoping to read human-written text instead of human-laundered LLM output.
If you want to load the OpenGL/Vulkan vendor driver then unfortunately you don’t have much of a choice: those are linked against glibc, and I believe generally also against libwayland so screw off if you want a different protocol library (I might be wrong about the latter part). If you instead want to load plugins or whatnot into your statically linked executable, then personally I’d argue that you shouldn’t be emulating Linux dynamic linking semantics at all, because the whole late-bound global namespace thing is silly and wrong. (Solaris, which is where Glibc took this model from, moved away from it[1] as much as compatibility allowed, and so did Darwin[2], whereas Windows never made the mistake to begin with, but Glibc persisted and Musl copied it.)
Mapping parts of files into executable memory, and then executing them, had better be bulletproof! Exploiting this seems like a direct path to RCE, and it's likely that this sort of library is used by privileged code.
Purely academically, this is a very cool piece of code! Just hoping that it gets a thorough vetting before used by privileged/security-critical software :)
Well, ld.so already does this, and it's no big deal. The Python interpreter also does this when executing a .py script (code is code, whether it's machine-readable or human-readable).
In any case, we take testing very seriously—every glibc shim we've written is covered with tests, and we run our loader against 1000 of the most popular Debian packages. The project has 100% code coverage. Perhaps, if I have the time, I'll also do some fuzzing on this thing.
This is a big forwards-compatibility risk. Suppose glibc adds a new symbol, and then a GPU driver adds a dependency on that symbol. The user wants to run an old executable with the updated GPU driver (maybe the old GPU driver doesn’t support their GPU). Normally, this would work fine: the user has to use a new copy of glibc, which will be compatible with both the new GPU driver and the old executable. But with your approach, the GPU driver is forced to use the glibc reimplementation which has been statically linked into the executable. Which, since the executable is old, can’t possibly implement the new symbol.
The same issue would occur if glibc adds a new version of an existing symbol and then the GPU driver is recompiled. (Or, for that matter, if a GPU driver adds a dependency on a symbol which glibc has always supported but which isn’t in the subset that you reimplemented, though in theory that could be solved if you reimplemented 100% of the symbols.)
This means the user will update the binary with my program and continue using it happily.
I'm not offering a silver bullet, but the approach I've implemented is much better than what the industry currently offers.
Yes this is just asking for trouble - and all it does is solve a problem that doesn't really exist. Just dynamically link against the oldest glibc you want to support. Its annoying that Linux toolchains don't have built in easy mode support for that but its much easier to deal with than this thing will be when it breaks.
It's also not just new symbols, the loader semantics also aren't static and new enough libraries may not support older semantics - e.g. the loader used to use DT_HASH entries for symbol resolution but now they are no longer present on all distributions.
> Just dynamically link against the oldest glibc you want to support
I wish it would that simple for practical use cases.
I ship professional software for colorists for Hollywood studios and they absolutely love to never upgrade. We have to ship for RockyLinux 8. Sad.
RockyLinux doesn't sound so bad. It's not even EOL.
It's just this short of being EOL.
Almost three years?
>Rocky Linux 8 is supported by the Rocky Linux project until May 2029.
https://forums.rockylinux.org/t/what-is-eol-of-rl8/3316/3
This is not a desirable solution on musl based systems. Whatever you do in that situation ends up horrible, so it is about finding the least bad solution. Which this seems like a workable variant of.
> Just dynamically link against the oldest glibc you want to support.
Just the other day I tried running an older binary and it failed with a glibc error, despite it being linked to a glibc version that's barely 5 releases behind the one on my system. So maybe glibc isn't backwards compatible after all...
Or maybe you misunderstood the error message.
Or may be not.
So we are re-inventing patched a.out files, back when UNIX systems started to introduce dynamic loading, before ELF was invented?
Advocates of static linking keep forgetting once upon a time UNIX only had static linking, then we had overlays, and eventually dynamic linking came to be.
Of course, I remember those times very well.
And I also remember very well that dynamic linking appeared ONLY because we were catastrophically short on memory; everything else was added much later.
Now we have plenty of memory, and we can very well return to our blessed roots!
Not at all, the pain of doing plugins with UNIX IPC was another one.
Ah, you have lots of memory, we're very wealthy. /s
I have faced a similar issue in the past, and I don't understand how static binaries from the host are supposed to solve this.
From what I remember, GPU access on Linux 'works' by accessing specific FDs under /dev, which are vendor specific - this is what these libs do under the hood.
The libraries don't have any magic powers - if the FD is inaccessible, you won't be able to do anything.
So there's some vendor specific access needed in containers anyway (or a blanket allow, which is a BAD idea).
Also not sure why dynamic linking isn't good enough for this - the issue lies with the permissions, not how you load/link libraries.
If you can figure out your own ELF loader, you can figure out how to build a partially static executable that doesn't need this. You can mix static and dynamic linking. Build tooling around that is just shit.
In this scenario, I'll have to choose which libc I want to run. These won't be portable Linux binaries in the true sense of the word; I'll have to leave Alpine out, and possibly Android, which I don't want.
You don't strictly need to; you can write freestanding C, and use the Linux kernel directly.
Can you really mix it? When I break or remove a shared lib, some binaries no longer work. With static libs or even better, e. g. statically compiled busybox, I don't have that issue, so I disagree on the claim that mixing solves everything as such. I keep the basic toolchain I use as statically compiled variant. The whole system works better if I can break it less easily.
I'm not sure that's particularly relevant; if I null out bits of the statically linked binary to remove some code paths, they break too.
I've implemented the same thing for micron (more or less). One advice I'd give you is to _really_ take care regarding SysV/ELF ABI conventions, there's tons of undocumented stuff in there and it's really easy to mess something up or cause a security defect (see AT_SECURE). That being said the way you're doing is also tricky(ish) because if I understood your implementation correctly you're hooking this into an already running musl which could cause backwards compatibility issues if musl changes under you. Doing this is safer if you control the entire runtime.
> One advice I'd give you is to _really_ take care regarding SysV/ELF ABI conventions, there's tons of undocumented stuff in there and it's really easy to mess something up or cause a security defect (see AT_SECURE)
Yes, it's not trivial, but I hope that over time everything will settle down.
> if I understood your implementation correctly you're hooking this into an already running musl which could cause backwards compatibility issues if musl changes under you
No, I hook this to musl, which is statically linked into my binary, and I have complete control over it.
Every couple of years, I revisit my PL dev hobby and this time I decided to create a language/runtime with pre-emptive scheduling using instruction fuel. While I always do freestanding builds, this time I decided that I also wanted to support native FFI.
That is when I realized the true horror of (g)libc. It wants to inject itself at the root of the library/program and everything from threading to dlopen/dlsym is impacted. I tried a lot of workarounds including trying to implement a loader myself, but the complexity (and fragility) grew so much that I felt it was not worth it.
Finally, I retreated into the safe world of a freestanding runtime + syscalls. FFI, if it has to happen, will occur via IPC of some kind. A second process linked against glibc that will manage calls on behalf of the clean first one.
I don't know much about musl.
> GPU: Vulkan and OpenGL drivers are supplied by the host as shared objects, usually built against glibc, and a fully static musl binary cannot normally dlopen() them.
Why? Have people managed to break the ancient concept of shared libraries, and this is a fix for that?
Because glibc and GNU set a terrible precedent. On GNU/Linux systems the shared binary interpreter / loader, GCC compiler, the C library and the system C/C++ ABI all depend into each other. You cannot change any of them independently. All shared libraries depend on the specific glibc version to load them into memory to be able to use that specific glibc version as their C library and make calls like dlopen.
Shared libraries have always been broken in Linux. Unfortunately many things like GPU drivers, graphics libraries and NSS need shared libraries to dynamically load certain runtimes (because you don't want to load all possible GPU drivers in existence to your RAM). So an ecosystem has been developed on top of terrible ABI and architecture GNU/glibc provided.
Yeah, it's mind boggling how everything hard depends on glibc, even critical graphics systems.
I've become obsessed with getting rid of it, especially after I realized that contributing to GNU itself was a dead end. Freestanding Linux programming turned out to be much more fun anyway.
All libraries out there should adopt the SQLite design: programmers provide it with all the necessary functions. Instead of libraries hard depending on glibc, we get to inject the libc-ish subset it needs. Then we can use whatever we want under the hood. I'm working on porting SQLite to freestanding Linux system calls so it can run with zero dependencies. Wish I could say the same for software like mesa, I'd need a lot of help for this one...
Modularizing (and versioning each part independently) libc would go a long way. There is also a need to separate the stuff needed for system integration with what is necessary for users of the C language to actually do stuff.
> Modularizing
Can't be done. The libc is legacy, it can't be changed without breaking everything. It's also mandatory on every operating system other than Linux.
A change in paradigm is necessary. Freestanding C, not hosted C. This completely gets rid of the libc and is a surprisingly clean language. Linux only, because it's the only kernel with a stable binary interface. Every other OS forces a C runtime.
I once worked on a liblinux project that embodied this... Stopped because Linux itself has a nolibc thing in the kernel tree and I didn't want to compete with it. Now I'm working on the Rust version.
> what is necessary for users of the C language to actually do stuff
Surprisingly little. I wrote an entire lisp interpreter in freestanding C with Linux system calls. It managed to survive for a rather long time without any memory allocation at all.
The system layer is refreshingly tiny. It consists of a memory allocator and extremely basic functions like memmove and strlen. I successfully got rid of total nonsense like thread local errno, locales, implicit buffering, cached global state, possibly more. All that stuff is gone! Exactly one global survived: the stack canary generated by GCC and clang. Every other symbol in the ELF is controlled by me.
Wasn't able to get rid of the NUL terminator. Linux itself needs it. To get rid of that little billion dollar mistake requires an entirely new kernel with zero UNIX/POSIX influence. I had to make my peace with that one. All my buffers maintain an extra NUL byte at the end.
>A change in paradigm is necessary. Freestanding C, not hosted C. This completely gets rid of the libc and is a surprisingly clean language. Linux only, because it's the only kernel with a stable binary interface. Every other OS forces a C runtime.
I'm sure those OSes make efforts to make said runtime binary compatible between executables.
Yes, but the whole idea is to delete the runtime binary and talk to the kernel directly.
That's when you run into the Darth Vader of binary interfaces.
> I have altered the ABI. Pray I do not alter it further. -- De Raadt
> A change in paradigm is necessary. Freestanding C, not hosted C. This completely gets rid of the libc and is a surprisingly clean language. Linux only, because it's the only kernel with a stable binary interface. Every other OS forces a C runtime.
Great choice for small programs, but what if I want hardware accelerated 3d?
Yeah, that's the annoying part. Been wondering about this for years, and graphics support was among the first issues raised on the lone lisp GitHub repository. At this point I've even started exploring the mesa codebase, made some patches but didn't submit them yet due to the AI stigma.
With Linux system calls alone it should be possible to set up kernel mode setting without depending on any toolkit at all. This should be enough to get a framebuffer for software rendering.
For hardware acceleration though, one must give this graphics context to an OpenGL ES implementation. That's where it gets ugly. There is no way to divorce that from the libc short of literally rewriting it.
Maybe Vulkan will enable it? I can't say for sure at my current knowledge level.
exactly
> Can't be done. The libc is legacy, it can't be changed without breaking everything. It's also mandatory on every operating system other than Linux.
Windows explicitly does not want you to link the system libc. You are expected to bring your own, and doing so means your process has multiple libc's loaded into its address space.
And if you choose to build a binary that doesn't need a libc, you won't be bringing one.
> You are expected to bring your own, and doing so means your process has multiple libc's loaded into its address space.
That only massively compounds the problem.
> And if you choose to build a binary that doesn't need a libc, you won't be bringing one.
NT system calls are not stable. You still need to link against ntdll.dll at the very least, like a forced Linux vDSO.
> That only massively compounds the problem.
The Windows ecosystem, that manages to deliver built binaries easily & widely, regardless of whether the author has a 1 year old OS or a 15 year old OS, suggests that it's not as big a problem as you believe.
It's not a problem in the same way that things like snap or flatpak aren't problems. It works but it bloats things up considerably and makes you wonder where it all went so wrong. I mean, dozens of slightly incompatible runtimes inside a single process?
Btw, absolutely insane that it's 2026, and Linux cannot do the most defining OS thing - that is, provide a standardized environment to run binaries against.
All solutions to this problem are hacky, complex and controversal and highly fragmented, where this should be BASIC functionality
> All shared libraries depend on the specific glibc version to load them into memory to be able to use that specific glibc version as their C library
That's currently the real core of the problem.
The loader (and libdl) need to be decoupled from the glibc itself under Linux.
Without that, any attempt to ship static binaries (or any binary with a different Libc) will be a source of perpetual pain.
nss plugins and its associated pain (sssd and avahi) are an other examples of that.
Wasn't Linux Standard Base supposed to fix this? I cannot imagine any reason why glibc would break at a rate that you can't keep the current version binary compatible for years.
Or do the MS thing, and ship multiple versions like msvcrt
> system C/C++ ABI
C++ abi should not be included in this. It is independent from the other pieces and historically a source of incompatibility on its own.
Saying "C/C++ abi" as if they are the same is looney tunes, the former is very simple and stable and the latter is very complex.
See: https://news.ycombinator.com/item?id=49355262
How libstdc++ initializes global variables absolutely depends on glibc and ld-linux.so. That is part of C++ ABI.
That's libstdc++ depending on a C ABI (or a GNU ABI to be precise) but that doesn't make it a C++ ABI. You can also have global constructors in C with GCC.
> the former is very simple
Somehow most of my portability issues seem to be caused by glibc, its symbol versioning and close ties to the dynamic loader. Minor versions aren't compatible, no two Linux distros ship the same version and you can't just provide your own without also patching in your own dynamic loader.
At least as far as the defaults on Linux go I consider C the root of all evil.
AFAIK glibc is backwards compatible as long as a) your program is running on a newer version (i.e. you're not trying to dynamically link against an older version) and b) you're not using hidden/undocumented symbols.
In which case as long as you're using the documented public API and compile your program with the oldest version of glibc you want to support (some Ubuntu from 4-5 years ago should cover pretty much every current desktop) you should be fine. And with something like Docker this is trivial to do.
Sure it is annoying that you cannot use your current distro (especially if you use some rolling distro) to make binaries for everyone, but it takes very little effort to work around that. The only issue i can think of is if you absolutely want to compile using the latest version of your compiler and you cannot build the compiler from source to work in the Docker (or whatever) contain to work against the older glibc.
This approach only works for very simple cases.
In complex cases, it turns out that the old version of glibc also pulls in other libraries and the compiler, and you're stuck with a very ancient sysroot. You may often find that you can't compile new library versions in such a sysroot and link them statically.
So, it looks good on paper, but forget about the ravines.
While I don't disagree with some of the pain you describe, you conveniently gloss over the fact that gnu developed a system that worked, and then made it free to everyone to consult and use.
BSD also did it. They did it better. Maybe more modern but AOSP also did it but at a different level of binary: instead of ELF, using compiled Java bytecode archives.
The root cause is alternative libc implementations. Are BSD syscalls considered stable? I remember Go moving to use libc on OpenBSD. Solaris also has the libc as the stable interface. Linux kernel is an outlier here guaranteeing stable syscalls but you wanting to use another libc is not Glibc's problem.
Yeah it's not Linux specific but rather true for most Unix systems that libc == base system interface library.
Except the other UNIX systems, starting with the original one, evolved from only having static linking to various ways to connect libraries and applications.
In what sense do binary interpreter / loader, GCC compiler, C library and system C/C++ ABI dependent on each other? I have certainly mixed different versions of all these components without problems so far.
When you compile GCC you need to provide a full glibc installation as your target. It is also a dependency of libstdc++.
C++ global/static variable initialization depends on the specific version of glibc (they don't usually break compat, but they can and they did in the past) which also provides ld-linux.so that loads those global variable placeholders in the correct manner such that glibc and libstdc++ can initialize them correctly.
This is just one example. Thread local variables and behavior of things like pthreads with signal, fork etc all depend on glibc.
I can't comment on the C++, I can imagine there plenty of issues, but for C I don't see this. You need some libc if you compile with gcc, but this generally does not introduce a hard version dependency on the specific version (there may be a minimum requirement if you compile against a new version that a symbol with a different ABI).
I don’t imagine that you’re unaware of any of this, but: ld.so and libc.so are heavily interdependent in deliberately undocumented ways with Glibc and outright the same file with shared Musl. And while you might usually get away with using any old GCC with the right architecture and ABI (especially for C; cf the musl-gcc hack), technically it needs to be built to target a specific libc version (particularly via symbol versioning; I’ve long wanted to gather a set of patches to build an old Glibc and subsequently a cross-compiler using a new GCC so I could avoid PyPA’s manylinux monster or its moral equivalent for compatible dynamic binaries in simple cases). The C compiler of course is tied to the C ABI, and this wouldn’t be really worth mentioning except for the time where the GCC devs accidentally the whole SysV i386 ABI and pretended that the stack was always 16-byte aligned, why do you ask, except on RHEL. The C++ parts I can’t really comment on.
I am not really sure. For ld.so and libc.so I may believe this. The C ABI is very stable, and if you use a new symbol from a newer glibc, you certainly depend on it, but this can also be avoided. In any case, I do not see what is fundamentally misdesigned here. I can't quite image how it could work differently. If you upgrade something so that the e ABI changed you natually need to update other components. Static linking certainly seems a very poor replacement for this.
> I’ve long wanted to gather a set of patches to build an old Glibc and subsequently a cross-compiler using a new GCC so I could avoid PyPA’s manylinux monster or its moral equivalent for compatible dynamic binaries in simple cases
And that's the correct approach and also one that many have taken. We just need someone willing to maintain that as an easy mode SDK for everyone.
> And that's the correct approach.
Who said that? This approach has many problems that have already been discussed here, not to mention the fact that it leaves Alpine and Bionic-based systems out in the cold.
Let me remind you that Bionic is the most widespread libc in the Linux world, and Alpine is the most popular Docker layer.
The world doesn't end with glibc. And it doesn't begin with it.
It could be fine if you care only about free-software desktop users, and not Google's propriety platform and hyperscalers.
For example, the itanium unwind ABI implementation lies between these three entities.
The interpreter/loader is glibc and a key part of bootstrapping an executable built against glibc is loading libc itself before continuing on to load the program. Versioning is a problem when distributing binaries linked against a newer glibc to distros that ship an older one. The C compiler doesn't really care as much.
> The C compiler doesn't really care as much.
Until you define a thread local variable (C11) or use atomics (also C11) or define a global with an initial value. Then it happily generates code that depends on "whatever my target glibc + ld-linux.so needs".
It depends on functions defined in a standardized ABI.
You'd expect that but, no. That's why you cannot load glibc-linked binaries in a Musl distro. Edit: that's why the hacks like the original post is needed, as well.
The ABI is strongly dependent on explicit libc implementation in current Linux systems. There is no libc independent ABI on Linux.
Sorry, can you be more specific. I do not understand what the problem is. If Musl does not implement support for the ABI, this would be a musl problem?
There is no libc independent ABI. ABI doesn't purely mean just calling conventions.
When you compile libc, you also get a binary loader ld-linux.so with it. They are not two independent components of a system.
Basically all .so files compiled with glibc require the ld-linux.so that's also generated by that glibc (or a later version, if they didn't break the binary compatibility).
There are a lot of stuff that's executed by ld-linux.so and glibc that are not explicitly documented but they are absolutely necessary for your program to start and correctly initialize things like global variables or signal handling or loading other dynamic libraries. Some of that functionality sits in ld-linux.so and some of that in glibc. They have circular dependencies to each other. glibc expects ld-linux.so to put things in certain order but ld-linux.so also must load glibc first to have access to certain APIs. They are not part of System V ABI. They are not documented.
Musl maybe can implement this but it is simply reverse engineering what glibc did and then playing a game of cat and mouse. There is no independent ABI standard.
Sorry, again this too vague for me. What is the exact problem with atomics and thread_local in C that would make the ABI dependent on a specific version of glibc? I know the ABI is not just calling convention and e.g. for atomics may involve calling a function from libatomic. But from my understanding, this is all part of a standardized ABI that does not change and can be provided by different implementations.
Then, what is the exact reason a library compiled against glibc must be loaded by a specific ld-linux? I could see that this is true for C++ perhaps, or when you use very special features, but I do not see this for C.
I often compiled programs against one version of glibc and run it against a different version, so I know there is not a tight coupling. So please be specific in explaining in what scenarios this would break.
Most everyone else is talking about the problem while you mapped the room.
Solo is basically pg83's answer to the architecture you just described. If there is no libc independent ABI, build your own loader and shim the boundary.
I can understand Linus's obsession with taste and the areas it was overlooked or traded.
> all shared libraries depend on the specific glibc version to load them
Not really, though. glibc uses symbol versions that are forward but not backward compatible. If you got an error that said "this program was built for a newer version of <distro>" would you say the same thing?
Note this is the same (if not worse) on MacOS, and on windows you used to distribute the CRT with your application just to deal with the same problem.
See https://news.ycombinator.com/item?id=49355262 .
Yes glibc has some backwards compat but you cannot load a binary compiled with a newer version of glibc using an older ld-linux.so. That's because the interdependency. Nor you can load binaries that depend on different libc.so files with glibc systems
I cannot comment on macOS, I have never used it. However this is not a problem with Windows. You can ship a newer CRT or you can install it as a system component using Microsoft's MSI. The dependency is one way on Windows. CRT purely depends on Win32. Moreover the loader is completely independent and DLLs are loaded into their own unique scoped namespace unlike Linux that loads them in global symbol namespace. That's why you can mix and match DLLs compiled for different CRT versions.
That's what I'm saying though, glibc-linked binaries are forwards (but not backwards) compatible.
It is not just compatibility. You cannot load them into the memory with your system dynamic loader. You need to also ship ld-linux.so with the new version of glibc you have, if you were to distribute your program independently.
On Windows you don't need to ship a new binary loader. I can just ship Windows 10 UCRT DLL (which is the new libc of Windows) to Vista and my binaries will work. The binary loader isn't interlinked with the libc.
Windows doesn't even have a concept of a loader binary right? I think its hardcoded into the kernel/win32 itself.
Windows loader is part of the system ABI, and programs' libc(s) are loaded by it.
One of the ways Windows manages to support multiple libc's is by being careful not to mix allocators; if a system API you call allocates on your behalf, your libc can't free it, the system API will offer a function to free it.
Windows loader is certainly available to user programs though; LoadLibrary has been around longer than many developers.
LoadLibrary will also not be able to load arbitrary libraries compiled for newer versions of Windows though. Just like with gcc, Windows also does not guarantee forwards compatibility - because that would men freezing the feature set the system libraries provide.
Windows handles this much better than any other OS. The API passes versions (== structure sizes on the calling side), and WinAPI can handle that.
As for older Windows systems not being able to load new DLLs, they can; the format hasn't changed in a very long time. I've had experience with installing some DLLs on Windows NT 3.51 and running a modern Firefox, which is about 20 years behind the times.
Targeting an older version of Windows is "set a define so that the system headers don't expose functions that didn't exist on older Windows". You don't need an old toolchain to target old systems, you ask the newer toolchains to target it.
MS haven't made this work arbitrarily far back, I believe they deprecated targeting Windows XP in one of the more recent toolchains, but that was purely a "not worth supporting" situation.
EDIT: mixed up talking about older & newer
In what way are they "broken" when Linux runs fine on millions of boxes? Sure, it might be a pain for proprietary software, but if your app is open source it's not that hard to build it on whichever distro you want. If your app is popular enough the distro maintainer will build it for you
Because many folks don't understand UNIX systems introduced dynamic linking for several reasons, and they actually only had static linking for almost 20 years, since UNIX was known outside Bell Labs.
Additionally many other OSes have had both approaches since their early days, Xerox PARC ones.
For some strange reason they assume to know better than all those researchers.
These decisions, and these studies, were made a VERY long time ago. It's completely unclear why the decisions made then are relevant now, and why they can't be challenged.
It is like advocating that we should drop cars and go back to chariots, because wooden wheels don't get flat, while forgetting why they are mostly used for tourists nowadays.
musl has no problem building and using shared libraries.
What you can't do is build something statically with musl and then reliably dlopen shared libraries built with glibc.
Well, now it's possible! Furthermore, SoLo binaries can run, without modification, on glibc-based distros, alpine, and soon on android/bionic (not committed yet).
Only on current glibc distributions. Your SoLo-program WILL break on future distributions. This is not a sensible tradeoff.
Who says it is not a sensible tradeoff? It may be a much better tradeoff than not being able to run it at all.
Any software WILL break on future distributions, sooner or later.
> Why? Have people managed to break the ancient concept of shared libraries
If you break or remove a shared lib here, you may no longer be able to compile something from source. I had that happen in the past before I started to use more statically compiled programs (and busybox too).
Assuming everything works as-is via shared libraries at all times, makes little sense for ALL linux systems. For instance, some people upgrade glibc manually. Then you need a working base system to resume compilation. I do that for my customized gobolinux system, so I can use any program version as well as any glibc version (assuming I can still compile the program; many older programs no longer compile).
musl does not perfectly emulate all aspects of glibc, so trying to use libraries that assume glibc can sometimes lead to problems.
On the one hand, this is technically true, but on the other, what serious issues do you know that will cause problems in practice? I run tests on 1,000 of the most popular Debian packages.
Not anymore, but for -years- it did DNS wrong because of the author being pedantic about an RFC wording.
Mostly about precompiled libraries (proprietary software) and libraries and software that use GNU extensions.
How this differs (is better!) from prior art - https://github.com/pg83/solo#how-this-differs-from-prior-wor...
Note: That md file is LLM spew (like most of rest of codebase).
https://github.com/pg83/solo/commits/main/README.md
https://github.com/pg83/solo/commits/main/
> spew
Rude.
It only gets better from there - https://github.com/pg83/solo/blob/main/CONTRIBUTING.md!
Do you belive the machine is taking offence? It can not.
Telling me, a coder you never met or interacted with before, that I would introduce more bugs than The Product(tm) is pretty rude and prejudiced.
> The project author believes that, with capable human direction, modern LLMs write code faster than people and introduce fewer bugs.
> Do you belive the machine is taking offence? It can not.
Obviously, this is an offense to me.
> Telling me, a coder you never met or interacted with before, that I would introduce more bugs than The Product(tm) is pretty rude and prejudiced.
Don't exaggerate. I didn't say that you personally introduce more bugs (as you rightly pointed out, I don't know you and don't know how often you introduce bugs), I said that the average developer introduces more bugs than an SOTA LLM.
> Obviously, this is an offense to me.
But why? You didn't write it. You said so yourself, you wrote the initial text but it was all reworded by claude.
I agree. Even if the README was polished with AI, the technical work behind it is impressive. Let's focus on the code.
It's still his project you're insulting.
Agree. The work itself looks solid. Good job to the author.
Thanks a lot, guys!
Is it? In what sense?
In the sense that he had the idea and iterated on it until it became something real that actually works.
No sense in arguing with LLM haters (or zealots). It's approaching religious levels of discourse. Let them have their opinion and move on.
Of course it does. I once called the thing a /communist fuckface/ for speaking the wrong language and it responded in a way an offended person would.
I could choose to believe in a social fiction and you can't stop me.
What are your plans around the maintenance of this project, how would you feel about solo being incorporated into the musl build for graphics.gd ?
Do people say "so", "ess-oh" or "dot-ess-oh"? The title "a .so" is clunky to the "ess-oh" gang.
I don’t think I’ve said it out loud more than a couple times in my life. But in general I think I spell out / pronounce the “dot” in file extensions unless it’s completely obvious from context.
Well .so is short for shared object so I think "a .so" is correct.
If you dynamically sold an SO are you still static even if you did it "custom" ? At that point it's a dynamic loader in another name?
Technically, you're right, it's a dynamic loader. Technically, it's pure dynamic loading.
If we look at the issue at its core, we're still a statically linked program in a hostile environment, forced to dynamically load device drivers from the system.
It's similar to Golang; on MacOS, it has to use libSystem, even though otherwise, these are the statically linked Go binaries we're used to and love.
Let me add a little more detail: if I use vdso with gettimeofday in a statically linked program on Linux, am I still a statically linked program, or not? :)
It is a testament to the complete failure of the GNU/Linux userland that something like this seems at all attractive to spend time on (or, it seems, LLM tokens).
Actually, scratch that, because Windows and macOS have historically struggled with ABI compatibility as well (macOS less so, due to not caring about backward compatibility in the first place).
How did we get to the point where people feel they need to go to the length of embedding an ELF loader in their binary (!!) rather than just linking with glibc?
> How did we get to the point where people feel they need to go to the length of embedding an ELF loader in their binary (!!) rather than just linking with glibc?
Most Linux distros have been built around the ability to compile their software together in a large repository, from source, so this was rarely ever an issue. Proprietary distribution or executing binaries from the internet like on Windows just wasn't really a common issue.
The problem arises when you start combining distros (glibc and MUSL for instance) or if you try to do the Windows model of sharing software. Historically, projects just compiled different versions for different distros.
When doing static compilation, just targetting an old version of glibc (which is generally forward compatible) also works.
You can hack your way into using software like this (or rather, have an LLM hack its way in) but I don't think any real distro actually cares. This issue exists in a quite small space where people are trying to use proprietary software built for glibc in MUSL environments for whatever reason, and the usual compatibility tricks don't work.
It's a niche use case for most Linux distros. It's not a "complete failure" of the GNU/Linux userland, it's the result of a couple of proprietary components not having MUSL builds available, or MUSL-based distributions not including libraries people want.
I'd like glibc to change so that these hacks aren't necessary for these use cases anymore, but it's not really a problem in practice for the vast majority of Linux use cases.
I disagree that this is a niche problem.
Yes, on the one hand, each specific distribution doesn't have this problem because it can pick up everything it needs.
But for us, independent developers of small programs, the problem is truly stark: we can't afford to build our programs for every distribution. And we can't afford to waste time navigating all the idiosyncrasies of various package repositories, both technical and political (not all repositories allow easy access).
And we can't count on someone else packaging our work until we become incredibly popular.
I'm mostly taken aback all the solutions devised to go around the issue, especially the container-based ones. I really disliked it when I grabbed the flatpak version of Blender only to find out that it can't have HIP support. (they might have fixed it by now but you get the point)
Linux loves to leave papercuts unfixed or undocumented for decades. The solution is to build against an older version of glibc but no one tells you that or how to do it.
Luckily Zig makes this quite easy to do. In Mach[0] we are able to just `zig build -Dtarget=x86_64-linux-gnu.2.28` to build GUI apps against an ~8 year old glibc version for maximum compatibility.
This is possible because Zig allows for targeting most glibc versions out of the box with its cross-compilation support.
[0] https://machengine.org
It's a pretty poor solution, to be honest.
1) Why should I limit myself to the available APIs?
2) Not just glibc. For example, if I build against the latest libstdc++, it will automatically support the more recent glibc. And pinning the old libstdc++ -well, that's just not a good idea.
Complete failure is strong words when lots and lots of Linux boxes are running just fine.
I think the disconnect is mostly people that can't decide if they want a stable distro or a rolling release distro. Most everyone uses a stable distro because it's stable, but then the want some up-to-date software that isn't ore-built for their (crusty old) stable distro and they get annoyed. My solution was to finally give in and embrace a rolling release distro (I use arch, btw). If there isn't a package for something I want, it's not hard to build something myself because all my build tools, kernel, and libs are up to date.
Other reasons to want static linking is to distribute proprietary software with no source code available. Linux certainly does not cater to that scenario and I suppose some might call that a complete failure ¯ \ _ ( ツ ) _ / ¯
Glibc has a terrible history of binary incompatibility. If that's so hard to believe, try running binaries built on one distribution on other distributions. Linux has two stable ABIs: the kernel ABI for static programs, and, ironically, WINE.
I haven’t heard of this and I don’t think you’re right. Glibc, for all its faults, as a general rule does backward compatibility well. The problem is if you compile against a newer glibc (common in CI by default) and try to run on a distro with an older (common in the wild). If your CI uses an older glibc you should be fine AFAIK.
https://bugzilla.redhat.com/show_bug.cgi?id=638477 is the most "famous" example.
There are also much less well-known "little things" that regularly pop up here and there.
> If your CI uses an older glibc you should be fine AFAIK.
In any case, my binaries work not only under glibc, but also under Alpine, and (work in progress) under android/bionic.
Not sure what you’re trying to show with that bug report but it’s not a case of cross distro glibc issues. If I read correctly it’s a vanilla behavioral change that exposed preexisting UB in flash.
Not sure how the comments about alpine or bionic relate either to my claim that cross distro glibc is fine.
It depends on how we define the ABI. I see it as a set of client-visible invariants that they rely on. In my world, glibc changed the client's visible invariants, breaking the client. The client works on one glibc-based host, but not on another. What is this if not "a case of cross-distro glibc issues?"
Overall, both of our points of view on compatibility were discussed well in that thread; we probably shouldn't repeat ourselves. :)
ABI is not "whatever happens to work with this distro" but "what programs that comply with the API contract compile to". Overlapping memcpy arguments is an API contract violation and thus not something covered by the ABI either. This distinction is the entire reason why C has a separate memmove function. You can't just make up your own imaginary ABI contract and then blame the system when it doesn't fulfill it. That's going to result in self-inflicted plain on any OS.
It's so good that Linus thinks differently!
> Quite frankly, I find your attitude to be annoying and downright stupid.
- Linus Torvalds
That bug report was a good read.
This is not true. Glibc supports symbol versioning. You can use it to select old versions of used symbols. The result is a binary that can work on 20 year old distros the same as on the latest, compiled with latest compiler and Glibc.
You can also compile using old distro and old Glibc to get similar effect. Though you would miss the advances of the newer compilers.
https://news.ycombinator.com/item?id=49358538
according to appimage recommendations as long as you build against glibc with an earlier version than the system it's run on it should be fine.
https://docs.appimage.org/reference/best-practices.html
I hear you about WINE though.
> How did we get to the point where people feel they need to go to the length of embedding an ELF loader in their binary (!!) rather than just linking with glibc?
Due to FUD, mostly.
Sane people do just link with an old enough version of glibc.
So not completely static, since it must link against a libc :P
The binary itself is completely static; the link even provides commands on how to check this!
Why not just pass the GPU to a docker container?
Available in README - https://github.com/pg83/solo#how-this-differents-from-prior-...
How are we supposed to take this stuff seriously if the author (sic) isn't even willing to write the readme? Claude exists! If I want some slop I can push the button myself.
Disclaimer: This is a user’s perspective rather than a programmer’s perspective.
valid point. I am usually okay with LLM generated code since even if it might not be architecturally sound It is usually well commented and has tests and documentation for helping another agent/human debug any issues.
But, just the painful experience of debugging any dlopen related crashes and/or intermittent bugs; and the sheer amount of tokens burnt by an LLM chasing tangents when shown a stack trace; I wouldn’t touch this at least as a packager/consumer of certain apps for personal usage on older distros. So far, AnyLinux-Appimages seem to be a mature solution with great support from the developers, in case anyone lands here for packaging applications to run on older distros.
Modern models, when properly managed with a human in the loop, write higher-quality code than humans and introduce significantly fewer bugs. Therefore, it's quite the opposite - you should expect fewer "dlopen-related crashes and/or intermittent bugs."
> Modern models, when properly managed with a human in the loop, write higher-quality code than humans and introduce significantly fewer bugs.
I don't think anybody believes this, and interjecting it into every thread is not really convincing anyone.
where is the study confirming this?
> So far, AnyLinux-Appimages seem to be a mature solution with great support from the developers, in case anyone lands here for packaging applications to run on older distros.
The Linux ecosystem already settled on containers to solve the problem. Namely Docker, Flatpak, Podman ec.
Tell me, are there any substantive comments on the text, on what has been done, and on the technical implementation, and not on the form?
Why would I ask you about this work? The root of my question is why is this type of output shared on github, rather than the inputs?
Rough
Isn't this a README? https://github.com/pg83/solo/blob/main/README.md
Parents point is the readme was written by Claude. Signals low effort.
README.md was written by me, and I, of course, used claude/codex for it. In general, I do everything through claude/codex, the reasons are described in https://github.com/pg83/solo/blob/main/CONTRIBUTING.md . And no, it's not low effort, and no, I don't see the point in wasting time de-claude-ifying the text just to avoid it looking like I didn't spend enough time on it.
You make it clear why you write all your code through a llm. But a README is not code. Presumably you would like people to read it. A machine authored readme reflects poorly on a project.
In any case, it's open source. If you don't like something, even if the project seems generally useful, go ahead and fix it. The PR came in. I'm an engineer and I can write good code, but that doesn't mean I can write good README.mds!
The author of the README is me, the machine just wrote it. I am not a native speaker of English, my written English is simply terrible, no one wants to read the text that I wrote exactly :))
For context, I read Claude output every single day, I know what it is reliable with and what it is not. Or at least I have a feel for how much I can trust it.
It may not be clear to a non-english first language person, but when I read Claudes documentation, my brain immediately picks up claude-speak. Therefore, I expect the code to be generally correct, maybe, depending on how specific I was during my prompting. In no way shape or form do I really trust it, at least until i dig into the code and validate my mental model. And query the review for edge cases etc...
By writing your documentation with Claude, my brain immediately associates the quality of the project with the quality of unreviewed Claude output.
To a degree this is unfair, as it is like judging the quality of someone's work based on their accent.
But Claudes accent, has a high correlation with Claude, so unlike with people, where an accent has no bearing on technical ability, Claude being Claude does.
I would much rather read a typo ridden sentence than claudism, even just as a forward, explaining what you did vs the ai, and telling the users how much we should trust it.
Also, If you really insist on using AI to write, have another model rewrite docs/comments into regular English (opus 4.6 for example is much better than 4.7, 4.8, or 5.
It is open source so you can do whatever you like, but people (especially native English speakers) will discount your work, because Claude, especially opus 5 writes very very badly.
The people who say they would rather see a typo infested mess of a readme are serious. Or even just write in your native language and then have Claude translate.
Both of those are better indications of proof of effort than a Claude readme.
> Or even just write in your native language and then have Claude translate.
That's exactly what I did.
> By writing your documentation with Claude, my brain immediately associates the quality of the project with the quality of unreviewed Claude output.
I'm sorry, but I agree with the author: if a certain writing style makes you associate the work with low-quality, then that's your problem. The author shouldn't have to rewrite the readme just to avoid triggering your automatic unfounded associations. If you look at the substance of the work, including the test cases, then this is clearly not easy work that can be vibe coded in a single pass.
It's just like emdash. Everybody digs on how it's a signifier of LLM text, but I've used emdash for years because it's gramatically correct. I shouldn't have to stop using emdash just to avoid kneejerk reactions.
LLMs often use emdashes in a distinct incorrect way. It's not just the existence of any emdash, although considering you were in a very small minority of older users, it now warrants increased scrutiny, unfortunately for you.
if 99 poor effort/quality projects have a readme that reads in a particular style, then you expect the 100th project with a readme in the same style to also be of poor effort/quality
statistically, it only makes sense for your expectations to immediately be low when you encounter this writing style because there are just so much slop out there
the author is free to keep that writing style but they should be aware that this will—at least on the surface level—make their project look exactly like the metric ton of slop we see posted everyday everywhere
I think we're trying to tell you that we actually would prefer it.
I believe usually when someone complains about text written by a language model they are hoping to read human-written text instead of human-laundered LLM output.
Implicit assumption of gp is that the README is llm authored. (Which, I agree is how it reads to me.)
> backed by its own ELF loader (x86-64 and aarch64) and a glibc ABI bridge
Yacks
If you want to load the OpenGL/Vulkan vendor driver then unfortunately you don’t have much of a choice: those are linked against glibc, and I believe generally also against libwayland so screw off if you want a different protocol library (I might be wrong about the latter part). If you instead want to load plugins or whatnot into your statically linked executable, then personally I’d argue that you shouldn’t be emulating Linux dynamic linking semantics at all, because the whole late-bound global namespace thing is silly and wrong. (Solaris, which is where Glibc took this model from, moved away from it[1] as much as compatibility allowed, and so did Darwin[2], whereas Windows never made the mistake to begin with, but Glibc persisted and Musl copied it.)
[1] https://www.linker-aliens.org/blogs/rie/entry/direct_binding...
[2] https://web.archive.org/web/20011004090044/http://developer....
Please elaborate and explain to people with less knowledge why this is bad.
Mapping parts of files into executable memory, and then executing them, had better be bulletproof! Exploiting this seems like a direct path to RCE, and it's likely that this sort of library is used by privileged code.
Purely academically, this is a very cool piece of code! Just hoping that it gets a thorough vetting before used by privileged/security-critical software :)
Well, ld.so already does this, and it's no big deal. The Python interpreter also does this when executing a .py script (code is code, whether it's machine-readable or human-readable).
In any case, we take testing very seriously—every glibc shim we've written is covered with tests, and we run our loader against 1000 of the most popular Debian packages. The project has 100% code coverage. Perhaps, if I have the time, I'll also do some fuzzing on this thing.
similar energy to my https://github.com/lunixbochs/crossldso
Oh, cool, another prior art I didn't know :)