TarWriter writes badly formatted size field for a zero sized regular file

[sobczyk]

Description

TarWriter writes badly formatted size field for a zero sized regular file

Reproduction Steps

using System;
using System.IO;
using System.Linq;
using System.Formats.Tar;

MemoryStream emptyData = new(0);
MemoryStream output = new();
TarWriter archive = new(output, TarEntryFormat.Pax);
PaxTarEntry te = new(TarEntryType.RegularFile, "zero_size")
{DataStream = emptyData};
archive.WriteEntry(te);
// get the size[12] field, after two unrelated blocks[512] and name[100], mode[8], uid[8], gid[8]
var offset = 2*512+100+3*8;
var sizeBuffer=output.GetBuffer()[offset..(offset+12)];
// it's "00000000000\0"
byte[] expected = new byte[] {0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0};
Console.WriteLine($"Proper size field: {sizeBuffer.SequenceEqual(expected)}");

Expected behavior

get all "0" in field (0x32) followed by "\0"

Actual behavior

get binary zero

Regression?

No

Known Workarounds

None

Configuration

Affected: (any with TarWriter)

• .net 7
• .net 8

Other information

No response

[sobczyk]

This breaks unpacking by tar-rs alexcrichton/tar-rs#347

[ghost]

Tagging subscribers to this area: @dotnet/area-system-io-compression
See info in area-owners.md if you want to be subscribed.

Issue Details

---

Description

TarWriter writes badly formatted size field for a zero sized regular file

Reproduction Steps

using System;
using System.IO;
using System.Linq;
using System.Formats.Tar;

MemoryStream emptyData = new(0);
MemoryStream output = new();
TarWriter archive = new(output, TarEntryFormat.Pax);
PaxTarEntry te = new(TarEntryType.RegularFile, "zero_size")
{DataStream = emptyData};
archive.WriteEntry(te);
// get the size[12] field, after two unrelated blocks[512] and name[100], mode[8], uid[8], gid[8]
var offset = 2*512+100+3*8;
var sizeBuffer=output.GetBuffer()[offset..(offset+12)];
// it's "00000000000\0"
byte[] expected = new byte[] {0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0};
Console.WriteLine($"Proper size field: {sizeBuffer.SequenceEqual(expected)}");

Expected behavior

get all "0" in field (0x32) followed by "\0"

Actual behavior

get binary zero

Regression?

No

Known Workarounds

None

Configuration

Affected: (any with TarWriter)

• .net 7
• .net 8

Other information

No response

Author:	sobczyk
Assignees:	-
Labels:	area-System.IO.Compression, untriaged
Milestone:	-

[ghost]

Tagging subscribers to this area: @dotnet/area-system-formats-tar
See info in area-owners.md if you want to be subscribed.

Issue Details

---

Description

TarWriter writes badly formatted size field for a zero sized regular file

Reproduction Steps

using System;
using System.IO;
using System.Linq;
using System.Formats.Tar;

MemoryStream emptyData = new(0);
MemoryStream output = new();
TarWriter archive = new(output, TarEntryFormat.Pax);
PaxTarEntry te = new(TarEntryType.RegularFile, "zero_size")
{DataStream = emptyData};
archive.WriteEntry(te);
// get the size[12] field, after two unrelated blocks[512] and name[100], mode[8], uid[8], gid[8]
var offset = 2*512+100+3*8;
var sizeBuffer=output.GetBuffer()[offset..(offset+12)];
// it's "00000000000\0"
byte[] expected = new byte[] {0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0};
Console.WriteLine($"Proper size field: {sizeBuffer.SequenceEqual(expected)}");

Expected behavior

get all "0" in field (0x32) followed by "\0"

Actual behavior

get binary zero

Regression?

No

Known Workarounds

None

Configuration

Affected: (any with TarWriter)

• .net 7
• .net 8

Other information

No response

Author:	sobczyk
Assignees:	-
Labels:	untriaged, area-System.Formats.Tar
Milestone:	-

[danmoseley]

Thanks for the report. Any interest in offering a PR?

[sobczyk]

Ok, I'll try to fix this

[sobczyk]

@danmoseley is there a website that displays current tests status?
I get a lot of tar unit tests fails when building unmodified dotnet runtime,
and I have no idea whether these are expected.

[sobczyk]

It seems the main I checked out at the time failed some unit tests, after pulling today the tests pass.

[danmoseley]

It is unusual that any tests fail in main. It really would only happen if they were outer loop as those aren't validated on every PR.

[danmoseley]

Thanks for looking at it. It will need to include a test.

(This can either be a unit test or a real file test, but the latter need to get committed to a test file repo before a test here can consume them)