August 5, 2026
Fast builds, faster backlash
Faster Than Ninja
Build speed brag turns into a comment-section cage match over who’s really cheating
TLDR: A new benchmark said build2 got close to Ninja’s famous speed, and even looked better once Ninja’s setup time was counted. But the real fireworks came in the comments, where readers argued over misleading bragging, "cheating" for speed, and whether the comparison was fair at all.
A geeky speed test about software build tools somehow turned into pure comment-section theater. The post compared Ninja, a famously stripped-down tool for turning code into apps, with build2, a newer all-in-one rival trying to get as close as possible to Ninja’s lightning-fast reputation. On paper, Ninja finished the test project in 3.4 seconds, while build2 came in at 3.8 seconds. But then came the spicy twist: the author pointed out that Ninja usually arrives with a helper tool called CMake, and that setup step alone took 15.6 seconds. Suddenly the "fastest" tool didn’t look so untouchable.
That’s where the crowd piled in. One commenter instantly called out the victory lap, basically saying, uh, the numbers don’t prove your headline. Even better, Ninja’s own creator dropped into the thread like a celebrity cameo and admitted Ninja is fast partly because it’s "cheating" by simply refusing to do extra work. That confession only fueled the drama: is it genius minimalism, or speed by conveniently ignoring hard problems?
Meanwhile, other commenters launched side quests of their own. One missed the old-school simplicity of a plain --help menu, turning the whole debate into a mini rant about modern tools being annoying for regular humans. Another got suspicious about compression settings and benchmark tuning, while a skeptic flat-out questioned whether build2 and CMake were even doing the same job. The vibe was a mix of benchmark nerd fight, usability grumbling, and delighted gotcha-hunting—with the community treating every second like a scandal.
Key Points
- •The article evaluates build2 against Ninja as a benchmark for near-minimal build-system performance.
- •Xerces-C++ was chosen for the comparison because more complex projects made apples-to-apples benchmarking difficult.
- •Ninja completed a full from-scratch Xerces-C++ build in a mean of 3.429 seconds over 10 runs.
- •CMake took 15.6 seconds to generate the Ninja build files, making the article’s end-to-end CMake-plus-Ninja workflow about 19 seconds for a fresh dependency build.
- •build2 completed the same build in a mean of 3.808 seconds, and the article then examines disabling some build2 features, starting with its more precise source change tracking, to better match Ninja’s workload.