A utility for copying as much of a file as possible.

Usage: SalvageFile (action)
 Common Options:
 -h / --help       Print this help

 (s)alvage (source) (destination)
 Copies as much of a file as possible.
  Options:
  -b (number)       Size of byte buffer in bytes (default: 4096). Larger values can be used to speed up the copy proces
   s but at reduced accuracy
  -v                Print more information including % complete
  -q                Print errors only and suppress copy warning messages

 (d)iff (file one) (file two)
 Compares the blocks of each file and shows if they are different
  Options:
  -b (number)       Size blocks in bytes (default: 4096)
  -v                Show all block hashes
  -q                Only show blocks that differ

• This program tries to copy all the chunks of a file. Any chunks that fail are replaced with zeros.
• I've seen that sometimes using a smaller buffer (-b) results in worse recovery, so you may need to play with the setting to achieve best results.
• As far as I can tell there's no way to control the read timeout without using system-dependent calls (which i'm avoiding)
• This program uses the regular System.IO.File class to read / write files
• The source and destination file systems must support seeking (most do).

• Install dotnet. see Get started
• Build with dotnet build -p src

• Follow the CoreRT - build prerequesites
• Build with ./make.sh publish

1. Install dotnet on linux
2. run ./make.sh publish

• Using Visual Studio
  1. Launch the "x64 Native Tools Command Prompt for VS" command prompt
  2. Use msys bash to execute ./make.sh publish
• I've also had success using cmder which comes with bash
• Installing dotnet on WSL is rummored to work as well

• I don't have a Mac currently - TBD
• Presumably this is similar to linux. see Prerequisites for macOS

• Test source and destination filesystem for seek support (currently just throws an exception)
• Check if destination file already exists and prompt for overwrite
• Consider implementing a sequential read option to allow reading non-seekable filesystems
• Add a way to diff copies of a recovered file and see which chunks are different
  • maybe show % amount of 0's in the chunk
• Maybe make a chunk 'merge' tool where you select which chunks to keep from each file
• output which chunks were replaced with 0's