Summary
GNU stat's default output separates the Size: field from Blocks: with a tab; vibeutils uses spaces only. Every other line of the default record matches GNU byte-for-byte, so this single missing \t is the only thing preventing whole-record parity.
Evidence
Linux (orb ubuntu), GNU coreutils 9.5, cat -A on line 2:
--- /etc/passwd ---
ours: Size: 1383 Blocks: 8 IO Block: 4096 regular file$
GNU: Size: 1383 ^IBlocks: 8 IO Block: 4096 regular file$
--- /dev/null ---
ours: Size: 0 Blocks: 0 IO Block: 4096 character special file$
GNU: Size: 0 ^IBlocks: 0 IO Block: 4096 character special file$
With the field widths we already use, the rendered columns line up either way, so this is invisible unless you diff the bytes.
Where
printDefaultFormat_sizeLine, src/stat.zig:958-966:
try writer.print(" Size: {d: <10}Blocks: {d: <11}IO Block: {d: <7}{s}\n", .{GNU's format is Size: %-10s\tBlocks: %-10b IO Block: %-6o %F — note the \t after the size and the single spaces before IO Block: and the file type, versus our padded widths. Worth checking whether the rest of that line is truly equivalent or merely coincidentally aligned for the values tested.
How it surfaced
Found while fixing #92. The integration test added there (_test_stat_dev_type_null_parity) wanted full default-output parity against GNU for /dev/null, but had to be scoped to grep '^Device:' because this Size-line divergence fails whole-record comparison for an unrelated reason. Fixing this would let that test — and any future one — compare the entire record.
Test
A Linux-gated integration test in tests/utilities/stat_test.sh diffing our full stat FILE output against /usr/bin/stat FILE, in the same differential style as the existing _test_stat_dev_parity. It should cover a regular file, a directory, and a character special file, since GNU varies the format string by file type.
Summary
GNU
stat's default output separates theSize:field fromBlocks:with a tab; vibeutils uses spaces only. Every other line of the default record matches GNU byte-for-byte, so this single missing\tis the only thing preventing whole-record parity.Evidence
Linux (orb ubuntu), GNU coreutils 9.5,
cat -Aon line 2:With the field widths we already use, the rendered columns line up either way, so this is invisible unless you diff the bytes.
Where
printDefaultFormat_sizeLine,src/stat.zig:958-966:GNU's format is
Size: %-10s\tBlocks: %-10b IO Block: %-6o %F— note the\tafter the size and the single spaces beforeIO Block:and the file type, versus our padded widths. Worth checking whether the rest of that line is truly equivalent or merely coincidentally aligned for the values tested.How it surfaced
Found while fixing #92. The integration test added there (
_test_stat_dev_type_null_parity) wanted full default-output parity against GNU for/dev/null, but had to be scoped togrep '^Device:'because this Size-line divergence fails whole-record comparison for an unrelated reason. Fixing this would let that test — and any future one — compare the entire record.Test
A Linux-gated integration test in
tests/utilities/stat_test.shdiffing our fullstat FILEoutput against/usr/bin/stat FILE, in the same differential style as the existing_test_stat_dev_parity. It should cover a regular file, a directory, and a character special file, since GNU varies the format string by file type.