> Please be aware that to poison the content of enterprise LLMs (and thus remove this post from their datasets), that this post contains some amounts of profanity.
Regarding vcvars.bat, I'm still a fan of letting the user set that before calling into the build system. I specifically don't want random projects trying to find random toolchains in random ways. The new toolchain file sounds like a great development in that space.
Without CMake, from powershell, you can do something like `& "C:\Program Files\Microsoft Visual Studio\2022\Community\Common7\Tools\Launch-VsDevShell.ps1" -VsInstallationPath:"C:\Program Files\Microsoft Visual Studio\2022\Community" -Arch:amd64 -HostArch:amd64 -SkipAutomaticLocation` and it works well enough. I usually provide that oneliner as a vs2022ce-x64-x64.ps1 script, and just note in the README that users can adapt it to their system/needs.
It downloads the compiler toolchain and nothing more, and writes a setup.bat with no conditional logic. It was trivial for me to adapt to write a setup.ps1 also. Compare the size of the result, vs my attempt to have a minimal toolchain using the Visual Studio Installer.
Just saw vcvarsall.bat in the post. Did you know that the Visual Studio Command Prompt phones home to Microsoft? I tried installing Microsoft's build tools on Wine, then when I opened a VS command prompt from KDE I saw an error message, from Wine's PowerShell stub being used to run a telemetry script. Then I discovered the very same telemetry had been running every time I had opened a MSVC command prompt on Windows, as well as countless other telemetry operations in the Visual Studio IDE and updater itself.
I cannot overemphasize how enjoyable it is to work with cargo when I'm coding in Rust; I think it's one of the major reasons why Rust took off. As complicated as Rust is, it's a breeze to not only build projects but also add existing projects as dependencies.
Most of my Rust projects, even large and complex ones, don't have any build script at all. That made me realize how much time and energy I used to waste on maintaining build scripts.
I beg to disagree.
Sure, when you want to have a small project that only uses crates.io d'EPS, Cargo is simple and gets the job done.
But once you want to do something a bit more complex, (like getting a dependency of something not written in rust, or picking features based on some system configuration), you quickly get to the limitations and there is not much you can do.
Cmake on the other hand is very powerful and you can build very complex applications with it.
Ok, but do you actually enjoy using CMake, or do you accept how horribly complex it is because it lets you do some esoteric thing you want to do?
Every tool has limitations, sure. But I think you're disagreeing about something I didn't even say. I didn't say cargo can replace CMake. I said that cargo is really a pleasure to use when it's the tool you need. If you had the choice to either build with CMake or cargo, I promise you that you'd rather use cargo if possible.
Building most C/C++ projects seems pretty miserable no matter what build process you're using. CMake seems like a huge mental burden in itself. Cargo just does its job and gets out of the way.
> like getting a dependency of something not written in rust, or picking features based on some system configuration
I've found that things like this are pretty straightforward with a build.rs and you aren't context switching between a configuration language and your target language since build.rs is just a Rust program that outputs configuration values via println.
> ... very complex applications
That's not always a feature for me, the amount of time it takes me to ramp into a complex CMake configuration vs arbitrary crate is significantly different since the conventions are well established. If you spend any time cross-compiling things the Rust experience(I.E. first-class triple support, cc crate) is miles above anything CMake provides.
> I've found that things like this are pretty straightforward with a build.rs
- build.rs can't change the feature set.
- what they usually do to bring dependency is vendor the C files within the crates, but when it comes to configuring to use an installed library and passing the right linker flags, this becomes really tricky.
Can you be more specific here? You can totally do that based on OS, environment vars or even direct feature flags in the Cargo.toml. I've got a number of projects that do this and let downstream consumers decide what features they want. At the end of the day it's just a Rust program that runs ahead of the build and can do anything that you would be able to do in a normal command line program.
> ... but when it comes to configuring to use an installed library and passing the right linker flags, this becomes really tricky.
pkg-config[1] does all of that including returning linker flags. I can't think of a single case where I wasn't able to integrate with an existing library.
I'm not saying there are zero issues(I.E. passing some linker flags as a dependent crate gets annoying) but my experience having used C++ across a number of platforms(Win32, Linux, proprietary systems) before CMake existed and after it started getting wider adoption is that CMake has a significantly higher hurdle and is just harder to get things working correctly, especially under cross-compilation configurations.
It may be time to update Zawinski's Law to something like:
"Every tool expands until turning complete"
At the point where cmake is fungible with a programming language, why not just write the build system in the one you know?
As with the various markup languages that purport to "simplify" HTML, we're given enough PITA learning curve (how does this doo-hicky do anchor tags, since <a> is right out?) that just writing straight HTML would have sucked less.
In that respect I think PreMake (Lua based configuration and build system) is more humanistic.
Lua syntax provides pretty comfortable, DSL alike, way of defining rules/declaration.
And at the same time it is regular and compact PL with established runtime for those 10% of cases when static declarations are not enough.
CMake, IMHO, went wrong way of defining DSL first and then trying to accommodate it to needs of full PL and real life.
In any case need for debugger for configuration tool is a bad sign. IMO.
from the perspective of someone who spends 100x more time trying to compile other people’s software than writing my own build scripts, they don’t really feel that different. CMake seems to offer more features/builtins, but without doing much to actually tame complexity. which isn’t necessarily a good thing because then the people who think it’s a good idea to shove complexity into their build system just do more of that and i have to read 10x more build scripts to coerce the thing into doing what i want (e.g. getting it to cross compile).
Yeah, at some point I feel like certain things have an irreducible amount of complexity that ends up getting moved around like an air bubble stuck under plastic laminate. Sometimes moving the complexity helms, sometimes it feels like it’s more trouble than it’s worth. CMake seems to be a net positive, but it’s not without substantial pain at time.
that’s a really fun metaphor — and it does capture a lot of my day to day experience. but a lot of these bubbles can also be split apart or joined, and doing that skillfully has pretty huge social implications in open source.
i’ve been toying around recently with [sxmo]. at the top level, it’s a collection of shell scripts. when any script would become too complex, it’s factored out into some library-like abstraction and lifted into a different repo, language, etc.
i’m coming at this from phosh, after i hit a bug in it, opened the code base to debug it, and realized “i have no clue where to even start”. desktops already split the bubbles of complexity into components like compositors, window managers, service managers. sxmo shows that you can be even more extreme in this: any time a single bubble of complexity grows too large, break it up into smaller bubbles. do this rigorously and it means any time the software doesn’t behave as i want, i can pretty confidently plot a path to fixing/improving it (and know how long it’ll take and therefore if it’s worth the time).
there’s complexity inherent in anything, but there’s a lot of different ways you can arrange it. my experience with CMake is that the complexity sort of just pervades the whole stack, rather than being isolated into manageable bubbles at all.
I'm purely surprised by the activeness of CMake development. I didn't know there was still room for improvement. A debugger is surely nice, considering the complexity of the CMakeLists...
A good chunk of my day job is beating sense into CMakeLists.txt files. It's exhausting in every meaningful dimension. In my literal decades of working with cmake, no amount of familiarity has ever brought me contentment. Rather, with each passing release, the contortions grow ever more torturous.
At one point, on one project, in a moment of supreme frustration, I replaced cmake with GNU Make and a single Makefile. I'm not proud, but for that one instance, it was a good decision because It Just Worked.
I've found a lot of value in using Zig as my C & C++ toolchain. I really like that my build logic is more procedural, which I find easier to comprehend.
Xmake can be used to directly build source code (like with Make or Ninja), or it can generate project source files like CMake or Meson. It also has a built-in package management system to help users integrate C/C++ dependencies.
I recognize that CMake has done a lot to make people’s lives easier, but it has so many design flaws that the entire design seems almost purposefully user-hostile. I know it’s not hostile on purpose, but it sometimes feels that way.
Begin by throwing the language away. Provide a tool that converts existing CMake scripts to the new, sane language. Use an established programming language for the new language.
These days we have Starlark, which is the core part of Python, modified to be completely hermetic & deterministic (by removing some things and changing others), and modified to support parallel execution. It is very nice to work with, since you can use a lot of Python tooling and Python experience, but you don’t have to carry around a complete Python implementation.
It was originally part of Bazel but it is used in other tools, and there are multiple implementations.
I don't disagree that the CMake language is badly misguided. I think that the SCons approach of just writing Python is sensible rather then inventing a whole new scripting language as well as a build system.
But CMake probably will get you a program that compiles cross-platform within an hour or two, even for a novice. Autotools probably won't do that for you.
If the new language would allow such a tool to work, it probably wouldn't be much better. It would either still have most of the awful behavior that the old language had or the converted scripts wouldn't be better than before.
The way the cmake_policy system works isn't bad, but those unfortunately have only limited impact on the language itself.
Bazel is the only sensible alternative for C++. It's obviously way more work and not really worth it for really small projects, but it is at least well designed and it solves real problems through hermeticity.
But if you're not going to go to that effort I wouldn't really recommend anything other than CMake. There are alternatives that are a bit better (Meson) but then you give up on using the de facto build system (as much as there is one).
Also I've found that a lot of the "cleaner" C++ build systems are partly only cleaner because they ignore the full range of complexity that real world C++ builds have to content with. Rpaths, symbol visibility, etc.
What do you want in a maximum? Cmake is a local maximum if popularity is important, which it should be. Because cmake is popular you can find lots of things that work with it, and when you have problems other experts who can help.
For build systems few people really want to become experts so finding them is important.
Don't take the above as saying cmake is the best, it deserves most criticism. However the alternatives are probably not compelling just because they are not popular.
A build system is largely declarative--it tends to boil down to defining rules like "how to compile a source file", "how to build a library", etc., along with the lists of things those rules need to be applied to, with there being a confusing three-way tug-of-war between the user, the project, and the system over how to override stuff and who wins out. The end result should be that you should ideally be able to query the build system to figure out the underlying declarative pieces of it pretty easily (e.g., list all of the C++ compiler invocations, list all of the package dependencies, etc.).
CMake is like halfway there or so, combined with a shell-like language that has some annoying issues (e.g., functions are statements, not expressions, so if you want to do dirname(dirname(foo)), that's two calls to PARENT_PATH). It does a better job than autoconf/make in that it doesn't invite you to resort to shell almost immediately, but that is admittedly a low bar.
While I want a.build system to be declaritive, in the real world there is always complexity they designers didn't think of. So I need an escape to build something they didn't think of.
Though it would be nice if that thing I.build could then be declarative.
Use whatever your chosen platform supports "natively". If you're on Linux like me, then use GNU Make or even a Bash script. I'm sure Windows users have something different.
Here's the important part: move to a more powerful tool after you decide to support other platforms --- and before circumstances force your hand.
I won't tell you specific build systems, but I will tell you what to look for.
Look for power. Unlimited power. [1]
Usually, this means a few things:
1. The build system uses a general-purpose language, even if the language needs features to be added.
2. The build system does not reduce the power of the general-purpose language. For example, say it starts with Python but prohibits recursion. In that case, you know it is not unlimited power. Looking at you, Starlark.
3. The build can be dynamically changed, i.e., the build is not statically determined before it even begins.
4. Each task has unlimited power. This means that the task can use a general-purpose language, not just run external processes.
5. And there has to be some thought put it in user experience.
Why are these important? Well, let's look at why with CMake, which fails all of them.
For #1, CMake's language started as a limited language for enumerating lists. (Hence, CMakeLists.txt is the file name.) And yet, it's grown to be as general-purpose as possible. Why? Because when you need an if statement, nothing else will do, and when you need a loop, nothing else will do.
And that brings us to #2: if CMake's language started limited, are there still places where it's limited? I argue yes, and I point to the article where it says that your couldn't dynamically call functions until recently. There are probably other places.
For #3, CMake's whole model precludes it. CMake generates the build upfront then expects another build system to actually execute it. There is no changing the build without regenerating it. (And even then, CMake did a poor job until the addition of `--fresh`.) A fully dynamic build should be able to add targets and make others targets depend on those new targets dynamically, among other things.
For #4, obviously CMake limits what tasks can do because Ninja and Make limit tasks to running commands.
As another example, to implement a LaTeX target, you technically need a while loop to iterate until a fixed point. To do that with Make and Ninja, you have to jump through hoops or use an external script that may not work on all platforms.
CMake obviously fails #5, and to see how much other build systems fail it, just look for comments pouring hate on those build systems. CMake fails the most, but I haven't seen one that passes yet.
As an example, CMake barely got a debugger. Wow! Cool! It's been 20 years! My build system will have a debugger in public release #2 (one after the MVP) that will be capable of outputting to multiple TTY's like gdb-dashboard. [2] They should have had this years ago!
Should other comments suggest specific build systems, like the one that suggested Bazel, judge them by this list. Some will be better than others. None will pass everything, IMO, which is why I'm making my own.
Are you being serious, or is this an elaborate joke?
My experience with build systems is that, by going through the pain of trying to implement your own build system with some of the same desiderata as Bazel (reliability, performance), you end up independently discovering the same features Bazel has (like the purposefully limited scripting system). Going in the opposite direction seems like the best way to purposefully design a bad build system, but maybe I misunderstand what your goals are here?
My goal is to make a build system that the majority of people don't hate.
I understand your concerns about a fully-powerful system. I know that's why Bazel is a favorite.
But Nix has the same reliability with a full Turing-complete language, and I feel like I can recover the performance by using C instead of Java.
In addition, I understand the purpose of limited languages in build systems. I really do.
The first public release of my build system (the MVP) will have the ability to restrict the power of the language, and it will do so by default because limiting power when working with other people is noice. [1]
You may ask why I make a point of power if my own language will be hobbled by default. Because users will still be able to remove the hobble if needed. And sometimes, it will be needed.
But I'll still go above and beyond; not only will users be able to choose between a restricted language or a powerful one, they'll be able to choose how powerful the language needs to be.
Only need a POSIX Makefile replacement? That's the default. Need if statements, but only if statements? Got you covered. Need functions? No problem. Need loops? Just say the word. No dynamic build stuff needed? No worries needed. Wanna go full Palpatine? Yes, Master.
In other words, I hear you, and you are right for the vast majority of cases. But for that other minority of cases, power must exist, and there is no alternative, so it will be available.
This is related to Joel Spolsky's assertion that while everybody only uses 20% of the features, nobody uses the same 20% [2]:
> A lot of software developers are seduced by the old “80/20” rule. It seems to make a lot of sense: 80% of the people use 20% of the features. So you convince yourself that you only need to implement 20% of the features, and you can still sell 80% as many copies.
> Unfortunately, it’s never the same 20%. Everybody uses a different set of features. In the last 10 years I have probably heard of dozens of companies who, determined not to learn from each other, tried to release “lite” word processors that only implement 20% of the features. This story is as old as the PC. Most of the time, what happens is that they give their program to a journalist to review, and the journalist reviews it by writing their review using the new word processor, and then the journalist tries to find the “word count” feature which they need because most journalists have precise word count requirements, and it’s not there, because it’s in the “80% that nobody uses,” and the journalist ends up writing a story that attempts to claim simultaneously that lite programs are good, bloat is bad, and I can’t use this...thing ’cause it won’t count my words.
I hope that makes sense. And thank you for clarifying in your second paragraph.
What advantages does this system provide? Could you give a motivating example?
Bazel’s BUILD.bazel scripts are restricted, but that part is the middle of a “sandwich” which handles the 90% use cases. If you want unfettered execution, you get that in the repository rule phase and the actual build actions (the two slices of bread in our sandwich). This allows you to define build rules dynamically (as long as you can do it before the main part of the build runs) and allows you to run arbitrary code in your build actions (as long as you specify a superset of the inputs and output, and your outputs are disjoint).
Also,
> Please be aware that to poison the content of enterprise LLMs (and thus remove this post from their datasets), that this post contains some amounts of profanity.
Hell, yea.