1Git pack format 2=============== 3 4== pack-*.pack files have the following format: 5 6 - A header appears at the beginning and consists of the following: 7 8 4-byte signature: 9 The signature is: {'P', 'A', 'C', 'K'} 10 11 4-byte version number (network byte order): 12 Git currently accepts version number 2 or 3 but 13 generates version 2 only. 14 15 4-byte number of objects contained in the pack (network byte order) 16 17 Observation: we cannot have more than 4G versions ;-) and 18 more than 4G objects in a pack. 19 20 - The header is followed by number of object entries, each of 21 which looks like this: 22 23 (undeltified representation) 24 n-byte type and length (3-bit type, (n-1)*7+4-bit length) 25 compressed data 26 27 (deltified representation) 28 n-byte type and length (3-bit type, (n-1)*7+4-bit length) 29 20-byte base object name 30 compressed delta data 31 32 Observation: length of each object is encoded in a variable 33 length format and is not constrained to 32-bit or anything. 34 35 - The trailer records 20-byte SHA1 checksum of all of the above. 36 37== Original (version 1) pack-*.idx files have the following format: 38 39 - The header consists of 256 4-byte network byte order 40 integers. N-th entry of this table records the number of 41 objects in the corresponding pack, the first byte of whose 42 object name is less than or equal to N. This is called the 43 'first-level fan-out' table. 44 45 - The header is followed by sorted 24-byte entries, one entry 46 per object in the pack. Each entry is: 47 48 4-byte network byte order integer, recording where the 49 object is stored in the packfile as the offset from the 50 beginning. 51 52 20-byte object name. 53 54 - The file is concluded with a trailer: 55 56 A copy of the 20-byte SHA1 checksum at the end of 57 corresponding packfile. 58 59 20-byte SHA1-checksum of all of the above. 60 61Pack Idx file: 62 63 -- +--------------------------------+ 64fanout | fanout[0] = 2 (for example) |-. 65table +--------------------------------+ | 66 | fanout[1] | | 67 +--------------------------------+ | 68 | fanout[2] | | 69 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ | 70 | fanout[255] = total objects |---. 71 -- +--------------------------------+ | | 72main | offset | | | 73index | object name 00XXXXXXXXXXXXXXXX | | | 74table +--------------------------------+ | | 75 | offset | | | 76 | object name 00XXXXXXXXXXXXXXXX | | | 77 +--------------------------------+<+ | 78 .-| offset | | 79 | | object name 01XXXXXXXXXXXXXXXX | | 80 | +--------------------------------+ | 81 | | offset | | 82 | | object name 01XXXXXXXXXXXXXXXX | | 83 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ | 84 | | offset | | 85 | | object name FFXXXXXXXXXXXXXXXX | | 86 --| +--------------------------------+<--+ 87trailer | | packfile checksum | 88 | +--------------------------------+ 89 | | idxfile checksum | 90 | +--------------------------------+ 91 .-------. 92 | 93Pack file entry: <+ 94 95 packed object header: 96 1-byte size extension bit (MSB) 97 type (next 3 bit) 98 size0 (lower 4-bit) 99 n-byte sizeN (as long as MSB is set, each 7-bit) 100 size0..sizeN form 4+7+7+..+7 bit integer, size0 101 is the least significant part, and sizeN is the 102 most significant part. 103 packed object data: 104 If it is not DELTA, then deflated bytes (the size above 105 is the size before compression). 106 If it is REF_DELTA, then 107 20-byte base object name SHA1 (the size above is the 108 size of the delta data that follows). 109 delta data, deflated. 110 If it is OFS_DELTA, then 111 n-byte offset (see below) interpreted as a negative 112 offset from the type-byte of the header of the 113 ofs-delta entry (the size above is the size of 114 the delta data that follows). 115 delta data, deflated. 116 117 offset encoding: 118 n bytes with MSB set in all but the last one. 119 The offset is then the number constructed by 120 concatenating the lower 7 bit of each byte, and 121 for n >= 2 adding 2^7 + 2^14 + ... + 2^(7*(n-1)) 122 to the result. 123 124 125 126== Version 2 pack-*.idx files support packs larger than 4 GiB, and 127 have some other reorganizations. They have the format: 128 129 - A 4-byte magic number '\377tOc' which is an unreasonable 130 fanout[0] value. 131 132 - A 4-byte version number (= 2) 133 134 - A 256-entry fan-out table just like v1. 135 136 - A table of sorted 20-byte SHA1 object names. These are 137 packed together without offset values to reduce the cache 138 footprint of the binary search for a specific object name. 139 140 - A table of 4-byte CRC32 values of the packed object data. 141 This is new in v2 so compressed data can be copied directly 142 from pack to pack during repacking without undetected 143 data corruption. 144 145 - A table of 4-byte offset values (in network byte order). 146 These are usually 31-bit pack file offsets, but large 147 offsets are encoded as an index into the next table with 148 the msbit set. 149 150 - A table of 8-byte offset entries (empty for pack files less 151 than 2 GiB). Pack files are organized with heavily used 152 objects toward the front, so most object references should 153 not need to refer to this table. 154 155 - The same trailer as a v1 pack file: 156 157 A copy of the 20-byte SHA1 checksum at the end of 158 corresponding packfile. 159 160 20-byte SHA1-checksum of all of the above.