Wait, like, you don't understand, or you don't share their complaint? I don't really understand how you don't understand. If I make a change to ripgrep because I'm debugging its perf and need to therefore create a release build, it can take several seconds to rebuild. Compared to some other projects that probably sounds amazing, but it's still annoying enough to impact my flow state.
ripgrep is probably on the smallish side. It's not hard to get a lot bigger than that and have those incremental times also get correspondingly bigger.
And complaining about compile times doesn't mean compile times haven't improved.
I do understand some factors, but I never noticed it being like super slow to build.
My personal project takes seconds to compile, but fair enough it's small, but even bigger projects like a game in Bevy don't take that much to compile. Minute or two tops. About 30 seconds when incremental.
People complained of 10x slower perf. Essentially 15min build times.
Fact that older versions might be slower to compile fills another part of the puzzle.
That and fact I have a 24 hyper thread monster of CPU.
30 seconds isn't incremental, that is way too long.
I work on a large'ish C++ project and incremental is generally 1-2 seconds.
Incremental must work in release builds(someone else said it only works in debug for Rust), although it is fine to disable link time optimizations as those are obviously kinda slow.
I'll echo Ygg2's comments. My previous job the minimum compile times were around 30 minutes so compile times under a minute feel like they're happening almost instantly. It's enough such that I don't need to break my thought process every time I compile.
Surely you can see how 1) it's all relative and 2) different people work differently. Like is this really so hard to understand? As far as I can tell, your comment is basically, "be grateful for what you have." But I am already. Yet I still want faster compile times because I think it will help me iterate more quickly.
I truly just do not see what is difficult to understand here.
First, compile times can differ wildly based on the code in question. Big projects can take minutes where hobby projects take second.
Also, people have vastly different work flows. Some people tend to slowly write a lot of code and compile rarely. Maybe they tend to have runtime tools to tweak things. Otherwise like to iterate really fast. Try a code change, see if the UI looks better or things run faster, and when you work like this even a compile time of 3 seconds can be a little bit annoying, and 30 seconds maddening.
It's less about "big projects" and more about "what features are used". It's entirely possible for a 10kloc project to take much more time to build than a 100kloc project. Proc macros, heavy generic use, and the like will drive compile time way up. It's like comparing a C++ project that is basically "C with classes" vs one that does really heavy template dances.
Notably, serde can drive up compile times a lot, which is why miniserde still exists and gets some use.
People are enabling serde codegen on every type, for no reason. That's it, that's the whole story. Those of us who don't do this will continue to read these "rustc is slow!!1!" posts and roll our eyes. Rustc isn't slow, serde is slow.
Completely agree, coming from a job where the C project I worked on took 30 minutes for basic software builds (you don't generally compile the code while writing it and spend a lot of time manually scanning looking for typos), Rust compile times are crazy fast.
Perhaps they were on old builds or some massive projects?