git/list[1] front-page[2] threads[3] people[4] search[5] about
 

Re: [PATCH 2/3] sha1_file: add the ability to parse objects in "pack file format"

From
PBPeter Baumann <peter.b.baumann@stud.informatik.uni-erlangen.de>
Date
Jul 12, 2006, 06:49 UTC
Message-ID
<slrneb96rd.dma.Peter.B.Baumann@xp.machine.xx>
In-Reply-To
<Pine.LNX.4.64.0607111656250.5623@g5.osdl.org>

On 2006-07-12, Linus Torvalds <torvalds@osdl.org> wrote: [...]

Show 12 quoted lines
> Anyway, I think this following patch replaces the old 2/3 and 3/3 (it 
> still depends on the original [1/3] cleanup.
>
> (It also renames and reverses the meaning of the config file option: it's 
> now "[core] LegacyHeaders = true" for using legacy headers.)
>
> Not heavily tested, but seems ok.
>
> sf? Dscho? Can you check this thing out?
>
> 		Linus
> ----
[...]
Show 56 quoted lines
> diff --git a/sha1_file.c b/sha1_file.c
> index 8734d50..475b23d 100644
> --- a/sha1_file.c
> +++ b/sha1_file.c
> @@ -684,26 +684,74 @@ static void *map_sha1_file_internal(cons
>  	return map;
>  }
>  
> -static int unpack_sha1_header(z_stream *stream, void *map, unsigned long mapsize, void *buffer, unsigned long size)
> +static int unpack_sha1_header(z_stream *stream, unsigned char *map, unsigned long mapsize, void *buffer, unsigned long bufsiz)
>  {
> +	unsigned char c;
> +	unsigned int word, bits;
> +	unsigned long size;
> +	static const char *typename[8] = {
> +		NULL,	/* OBJ_EXT */
> +		"commit", "tree", "blob", "tag",
> +		NULL, NULL, NULL
> +	};
> +	const char *type;
> +
>  	/* Get the data stream */
>  	memset(stream, 0, sizeof(*stream));
>  	stream->next_in = map;
>  	stream->avail_in = mapsize;
>  	stream->next_out = buffer;
> -	stream->avail_out = size;
> +	stream->avail_out = bufsiz;
> +
> +	/*
> +	 * Is it a zlib-compressed buffer? If so, the first byte
> +	 * must be 0x78 (15-bit window size, deflated), and the
> +	 * first 16-bit word is evenly divisible by 31
> +	 */
> +	word = (map[0] << 8) + map[1];
> +	if (map[0] == 0x78 && !(word % 31)) {
> +		inflateInit(stream);
> +		return inflate(stream, 0);
> +	}
> +
> +	c = *map++;
> +	mapsize--;
> +	type = typename[(c >> 4) & 7];
> +	if (!type)
> +		return -1;
> +
> +	bits = 4;
> +	size = c & 0xf;
> +	while (!(c & 0x80)) {
> +		if (bits >= 8*sizeof(long))
> +			return -1;
> +		c = *map++;
> +		size += (c & 0x7f) << bits;
> +		bits += 7;
> +		mapsize--;
> +	}

This doesn't match the logic used in unpack_object_header, which is used in the packs:

static unsigned long unpack_object_header(struct packed_git *p, unsigned long offset,
        enum object_type *type, unsigned long *sizep)
{
	unsigned shift;
	unsigned char *pack, c;
	unsigned long size;
	if (offset >= p->pack_size)
		die("object offset outside of pack file");
	pack =  (unsigned char *) p->pack_base + offset;
	c = *pack++;
	offset++;
	*type = (c >> 4) & 7;
	size = c & 15;
	shift = 4;
	while (c & 0x80) {			<==========
		if (offset >= p->pack_size)
	        	die("object offset outside of pack file");
		c = *pack++;
		offset++;
		size += (c & 0x7f) << shift;
		shift += 7;
	}
	*sizep = size;				<==========
	return offset;
}
Show 22 quoted lines
> @@ -1414,6 +1462,49 @@ static int write_buffer(int fd, const vo
>  	return 0;
>  }
>  
> +static int write_binary_header(unsigned char *hdr, enum object_type type, unsigned long len)
> +{
> +	int hdr_len;
> +	unsigned char c;
> +
> +	c = (type << 4) | (len & 15);
> +	len >>= 4;
> +	hdr_len = 1;
> +	while (len) {
> +		*hdr++ = c;
> +		hdr_len++;
> +		c = (len & 0x7f);
> +		len >>= 7;
> +	}
> +	*hdr = c | 0x80;
> +	return hdr_len;
> +}
> +
Dito, but in this case see pack-objects.c
/*
 * The per-object header is a pretty dense thing, which is
 *  - first byte: low four bits are "size", then three bits of "type",
 *    and the high bit is "size continues".
 *  - each byte afterwards: low seven bits are size continuation,
 *    with the high bit being "size continues"
 */
static int encode_header(enum object_type type, unsigned long size, unsigned char *hdr)
{
        int n = 1;
        unsigned char c;
        if (type < OBJ_COMMIT || type > OBJ_DELTA)
                die("bad type %d", type);
        c = (type << 4) | (size & 15);
        size >>= 4;
        while (size) {
                *hdr++ = c | 0x80;	<=======
                c = size & 0x7f;
                size >>= 7;
                n++;
        }
        *hdr = c;			<=======
        return n;
}
-Peter Baumann
Previous: Linus TorvaldsNext: Junio C Hamano
Message 32 of 37 in “Revisiting large binary files issue.”
  1. Carl BaldwinJul 10, 2006
  2. Junio C HamanoJul 10, 2006
  3. Peter BaumannJul 11, 2006
  4. Linus TorvaldsJul 10, 2006
  5. [RFC]: Pack-file object format for individual objects (Was: Revisiting large binary files issue.)sf, Jul 11, 2006
  6. Linus TorvaldsJul 11, 2006
  7. sfJul 11, 2006
  8. Linus TorvaldsJul 11, 2006
  9. Linus TorvaldsJul 11, 2006
  10. Carl BaldwinJul 11, 2006
  11. Linus TorvaldsJul 11, 2006
  12. 1/3 Make the unpacked object header functions static to sha1_file.cLinus Torvalds, Jul 11, 2006
  13. 2/3 sha1_file: add the ability to parse objects in "pack file format"Linus Torvalds, Jul 11, 2006
  14. Johannes SchindelinJul 11, 2006
  15. Linus TorvaldsJul 11, 2006
  16. Johannes SchindelinJul 11, 2006
  17. Linus TorvaldsJul 11, 2006
  18. Johannes SchindelinJul 11, 2006
  19. Junio C HamanoJul 11, 2006
  20. sfJul 11, 2006
  21. Linus TorvaldsJul 11, 2006
  22. sfJul 11, 2006
  23. Junio C HamanoJul 11, 2006
  24. Linus TorvaldsJul 12, 2006
  25. Johannes SchindelinJul 12, 2006
  26. Linus TorvaldsJul 12, 2006
  27. Linus TorvaldsJul 12, 2006
  28. Junio C HamanoJul 12, 2006
  29. Linus TorvaldsJul 12, 2006
  30. Junio C HamanoJul 12, 2006
  31. Linus TorvaldsJul 12, 2006
  32. Peter BaumannJul 12, 2006
  33. Junio C HamanoJul 12, 2006
  34. Peter BaumannJul 12, 2006
  35. Linus TorvaldsJul 12, 2006
  36. Junio C HamanoJul 12, 2006
  37. 3/3 Enable the new binary header format for unpacked objectsLinus Torvalds, Jul 11, 2006

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.