Build Directories and Worktree Disk Usage

A build directory is a workspace for a C/C++ build system, not a Git concept.

Source Versus Build

Source/
├─ Camera.cpp
├─ Scene.cpp
└─ Main.cpp

build/
├─ *.obj
├─ *.lib
├─ *.dll
├─ *.exe
├─ *.pdb
├─ CMakeCache.txt
└─ generated files

The Build Process

flowchart LR
    S[.cpp Source] --> C[Compiler]
    C --> O[.obj / .o]
    O --> L[Linker]
    L --> E[.exe / .dll / .lib]

Incremental Build

The first build compiles many files.

Subsequent builds can regenerate only the outputs of modified files.

Modify Camera.cpp
↓
Regenerate Camera.obj
Reuse Scene.obj
Reuse UI.obj
↓
Link

Deleting the build directory usually turns the next build into a more expensive clean build.

Why Multiple Worktrees Use More Disk Space

Worktrees share the Git object database, but build outputs are usually generated separately for each worktree.

WT A
├─ Source
└─ build 40 GB

WT B
├─ Source
└─ build 42 GB

WT C
├─ Source
└─ build 38 GB

The main contributors are usually:

• .obj
• .pdb
• .lib
• generated code
• shader cache
• dependency build outputs

Strategies

1. Keep Each Worktree’s Build Directory Separate

Sharing a build directory is risky.

WT A → build/A
WT B → build/B

2. Delete Build Outputs for Inactive Worktrees

This does not affect source files or commits. Regenerate the outputs when needed.

3. Remove Merged Worktrees

git worktree remove ../worktrees/camera
git branch -d feature/camera

4. Share a Compiler Cache

Caches such as ccache / sccache can be shared.

flowchart TD
    A[WT A / build] --> C[Shared Compiler Cache]
    B[WT B / build] --> C
    D[WT C / build] --> C

Reuse previously compiled outputs while keeping the build directories themselves separate.

5. Share Dependency Caches

Download and binary caches from dependency managers such as vcpkg / Conan can also be shared.

A Practical Cleanup Policy

Active Worktree
→ Keep build outputs

Paused Worktree
→ Delete build outputs if needed

Merged Worktree
→ Remove the worktree + build outputs

Shared
→ compiler/dependency cache

This preserves the safety of parallel development while reducing disk usage and repeated build time.