Build dependency checker gets 54x faster with eBPF
Researchers at Waseda University developed mkcheck2, a tool that uses eBPF system-call tracing and incremental analysis to verify build dependencies. In tests across open-source projects, it cut dependency-error detection overhead by up to 99.7% and sped up analysis by as much as 54x. The work was published at the 2026 IEEE/ACM International Conference on Software Engineering.
Researchers Yuta Saito, Kazunori Sakamoto, and Hironori Washizaki at Waseda University created mkcheck2 to catch build dependency errors. Build systems such as Make, CMake, and Zig automate compilation and linking, but dependency specification issues cause over half of build errors in large projects.
Existing ptrace-based tracing slows builds because it suspends processes and switches contexts. mkcheck2 instead runs eBPF tracing inside the kernel, adding incremental analysis. On a 300-project Make corpus, mean analysis time per commit fell from 1,267.49 seconds to 23.56 seconds. The approach remains Linux-specific and faces limits with memory-mapped regions, dynamic libraries, network dependencies, and distributed builds.
Faster dependency verification could reduce waiting in continuous integration, letting developers catch build errors sooner. Open-source maintainers and companies relying on large Make/CMake projects may benefit from lower compute costs and quicker feedback. Because gains are Linux-specific, cross-platform teams may see uneven effects. The tool may also encourage more frequent dependency checks, potentially improving software reliability, though real-world adoption and broader build-system coverage remain uncertain.