lwext4, branch fix-mac Fork of lwext4 library for EXT2 filesystem manipulation https://git.carlh.net/cgit/lwext4/atom?h=fix-mac 2021-12-17T21:11:54Z Fix inexplicable change to dylib paths with new VM. 2021-12-17T21:11:54Z cah cah@ableton.com 2021-12-17T21:11:54Z urn:sha1:90e53d5f1e3eadd3534b6fa1b0e3404b3a7cae1e After re-creating the macOS build VM, suddenly liblwext4 ends up linked as @rpath/liblwext4.dylib which breaks the DoM make_dmg script. I can only guess this is due to me installing a newer CMake. Fix some overflows when making larger partitions (e.g. 2TB) (DoM bug #2021) 2021-05-28T09:14:11Z Carl Hetherington cth@carlh.net 2021-05-28T09:14:06Z urn:sha1:cce3730330bd5621a94ec813934ba4f41254910d Merge branch 'carl2-opt' into carl2 2021-03-12T22:24:01Z Carl Hetherington cth@carlh.net 2021-03-12T22:24:01Z urn:sha1:9d20ec504c40c79a63c3ffa14dc9152e0ba4ecae Apply various optimisations, tested only on macOS. Use posix open/read/write/close as it seems to be much faster than stdio. 2021-03-12T22:21:50Z Carl Hetherington cth@carlh.net 2021-03-12T21:09:57Z urn:sha1:51774b7c8dcb3968330b28d9368b1b959b873670 Add progress reporting for formatting. 2021-03-09T00:03:41Z Carl Hetherington cth@carlh.net 2021-03-08T22:42:09Z urn:sha1:abb212dd5fbb33168f583f36b67f29f659a91bb4 Don't initialise inode tables during format. 2021-03-09T00:03:41Z Carl Hetherington cth@carlh.net 2021-03-08T23:16:48Z urn:sha1:6d34064b3a4b07218b01f6cc7700ee2ce68a7542 Use writebytes when writing group descriptors. 2021-03-08T23:27:27Z Carl Hetherington cth@carlh.net 2021-03-08T23:24:39Z urn:sha1:d217f6e5082fab342267df3430ca6b9f2e399c93 This may be faster but I should test it. The code is cleaner and should be equivalent, as far as I can see. Write block/inode bitmaps and inode tables earlier. 2021-03-08T23:25:39Z Carl Hetherington cth@carlh.net 2021-03-08T23:21:56Z urn:sha1:b13cbdce2fe1c4790f0737a75ac49aa1b826eb66 Before we would write blank block/inode bitmap tables then fill them in later. This is a bit of a hack to set up enough of the data structures (I think) to allow the correct block/inode bitmaps to be written first time. This may not be worth the hack/risk - I haven't tested the speed improvement carefully. Ignore .ccls-cache 2021-03-08T23:25:39Z Carl Hetherington cth@carlh.net 2021-03-07T22:43:54Z urn:sha1:c1fc6c7e573c644013bf65ae5ef474e235466dc0 Write all block group headers sequentially. 2021-03-08T23:25:39Z Carl Hetherington cth@carlh.net 2021-03-04T19:53:18Z urn:sha1:c375ac35c728ebd99d794b3c2b64a3ecb07f27b1 Before this the code would write: headers for blocks 0-127 (without group descriptors) group descriptors for blocks 0-127 to all blocks on disk headers for blocks 128-255 group descriptors for blocks 128-255 to all blocks on disk ...