views:

345

answers:

2

I've been pouring over the format description and source code for the 7z archive format, but I'm still having trouble writing a valid container. I assume I can create an empty container... anyway here's my start:

std::ofstream ofs(archivename.c_str(), std::ios::binary|std::ios::trunc);

Byte signature[6] = {'7', 'z', 0xBC, 0xAF, 0x27, 0x1C};
Byte major = 0;
Byte minor = 3;

ofs.write((const char*)signature, 6);
ofs.write((const char*)major, 1);
ofs.write((const char*)minor, 1);

UInt64 offset = 0;
UInt64 size = 0;
UInt32 crc = 0;

ofs.write((const char*)offset, 4);
ofs.write((const char*)size, 8);
ofs.write((const char*)crc, 8);
ofs.write((const char*)CrcCalc(0, 0), 8);

ofs.close();

I think my main problem is a lack of understanding of std::ofstream::write(). Byte is an 'unsigned char', UInt64 & UInt32 are both 'unsigned long'.

UPDATE0: As everyone points out, it'd be a problem if I ran this on a big-endian machine. That's not the case here. Per Fredrik Janssen, I should be casting the address of the non-arrays. I should also mention that CrcCalc() is a function in the LZMA SDK. Adding & helps a bit, it's that first unsigned char[6] that's having some problems.

UPDATE1: Working code to get an empty archive file below.

static void SetUInt32(Byte *p, UInt32 d)
{
  for (int i = 0; i < 4; i++, d >>= 8)
    p[i] = (Byte)d;
}

static void SetUInt64(Byte *p, UInt64 d)
{
  for (int i = 0; i < 8; i++, d >>= 8)
    p[i] = (Byte)d;
}

void make_7z_archive()
{
  CrcGenerateTable();

  std::ofstream ofs(archivename.c_str(), std::ios::binary|std::ios::trunc);

  Byte signature[6] = {'7', 'z', 0xBC, 0xAF, 0x27, 0x1C};
  Byte major = 0;
  Byte minor = 3;

  ofs.write((const char*)signature, 6);
  ofs.write((const char*)&major, 1);
  ofs.write((const char*)&minor, 1);

  UInt64 offset = 0;
  UInt64 size = 0;
  UInt32 crc = 0;

  Byte buf[24];
  SetUInt64(buf + 4, offset);
  SetUInt64(buf + 12, size);
  SetUInt32(buf + 20, crc);
  SetUInt32(buf, CrcCalc(buf + 4, 20));

  ofs.write((const char*)buf, 24);

  ofs.close();
}

NOTE: CrcGenerateTable() and CrcCalc() are from the LZMA SDK.

+2  A: 

don't know the format of 7z, but I notice when you write down offset, size and crc that these will be written to the file in little-endian format (I assume you have a little-endian CPU).

Edit: An probably worse, you are missing the & before major, minor, offset, size and crc, i.e. you are casting the actual values to a pointer.

Fredrik Jansson
Just noticed the )
Terry Lorber
The missing ampersands was the problem. In addition, I stuffed the other values into a buffer array, following the example in the 7z source. Thanks for the extra eyes.
Terry Lorber
Just glad to help :) Good luck with the project!
Fredrik Jansson
A: 

Uh... I'm confused. It has an SDK... which you mention in your post... also the 7-zip sources are online. see also p7zip on SourceForge. I just glanced at the sources for p7zip and there are a bunch of files that start with "7z" that look like they would do the trick.

I haven't used the 7z format programattically myself (just through the command-line util / GUI) but why would you need to handle those low-level things yourself rather than through the SDK? (other than because of the LGPL licensing)

Jason S
There's no convenient library for encoding. There's an ANSI-C library for decoding, a pre-built encoder/decoder, then a bunch of CPP code for an entire client app, and a DLL (just a COM interface). I need to link statically, and didn't feel like figuring out how to build just the code I need.
Terry Lorber
Oops, the ANSI-C library does do encoding, but doesn't handle the archive format. I've downloaded p7zip... more examples of this code in action!
Terry Lorber