使用boost::filesystem的C++目录大小工具与du结果不符求助

Hey there, let's unpack why you're seeing this mismatch between your C++ tool and the du command—great catch that Nautilus aligns with your results, that's a key clue! Here are the most common reasons for this discrepancy:

1. Apparent Size vs. Disk Block Usage #

This is the #1 culprit.

• Your tool (and Nautilus) uses boost::filesystem::file_size(), which returns the actual logical size of the file (the number of bytes the file contains).
• The default du command reports the disk space actually occupied by the file. Since files are stored in fixed-size disk blocks (usually 4KB on most systems), even a 1-byte file will take up an entire block. So du counts that as 4KB, while your tool shows 1 byte.

To verify this, run du --apparent-size on the same directory—this makes du report logical size, and it should match your tool's output perfectly.

2. Sparse Files Handling #

If your directory contains sparse files (files with large empty sections marked as unallocated), this amplifies the above issue:

• file_size() returns the full logical size (e.g., 1GB for a sparse file that only uses 1MB of disk space).
• Default du returns the actual disk space used (1MB in this example). Again, du --apparent-size will match your results.

3. Symbolic Link Behavior #

Double-check how your tool handles symlinks:

• boost::filesystem::directory_iterator doesn't follow symbolic links by default. So your tool counts the symlink itself (a tiny file, usually a few dozen bytes) rather than the file/directory it points to.
• By default, du also doesn't follow symlinks—but if you accidentally ran du -L (which follows symlinks), that would cause a mismatch. Conversely, if your tool is configured to follow symlinks but du isn't, that would also explain the difference.

Debugging Steps to Pinpoint the Issue #

• Pick a small directory where the difference is obvious. Compare your tool's output with both du and du --apparent-size—this will immediately tell you if block size is the problem.
• For individual files, run stat <file>: look at Size (logical size, matches your tool) vs Blocks (disk blocks used, matches default du).
• If symlinks are suspect, test with a directory containing a symlink: check if your tool counts the symlink itself or follows it, then compare with du and du -L.

If you want your tool to match du's default behavior, you can calculate the disk block usage manually using boost::filesystem::space() to get the block size, then compute each file's occupied blocks as (file_size + block_size - 1) / block_size, then multiply by block size to get the total.

内容的提问来源于stack exchange，提问作者jelmew