How to get the uncompressed size of an LZMA2 file (.xz / liblzma)

Question

I'm looking for a way to get the uncompressed stream size of an LZMA2 / .xz file compressed with the xz utility.

I'm using liblzma from Windows/Linux for this task, so I guess I'm looking for some C/C++ API in liblzma that will do the trick.

Answer 1

I think I've found a solution.

This is a very crude code sample, but seems to work fine.

I'm assuming I have a do_mmap() function that maps the entire file as read-only into memory, and returns the total size mapped. This can naturally be adapted to use read/fread/ReadFile or any other File API.

extern size_t get_uncompressed_size(const char *filename)
{
   lzma_stream_flags stream_flags;
   int file_size;

   const uint8_t *data = (uint8_t *) do_mmap(filename, &file_size);

   // 12 is the size of the footer per the file-spec...
   const uint8_t *footer_ptr = data + file_size - 12;

   // Something is terribly wrong
   if (footer_ptr < data) {
     do_unmap((void *)data, file_size);
     return -1;
   }

   // Decode the footer, so we have the backward_size pointing to the index
   lzma_stream_footer_decode(&stream_flags, (const uint8_t *)footer_ptr);
   // This is the index pointer, where the size is ultimately stored...
   const uint8_t *index_ptr = footer_ptr - stream_flags.backward_size;
   // Allocate an index
   lzma_index *index = lzma_index_init(NULL);
   uint64_t memlimit;
   size_t in_pos = 0;
   // decode the index we calculated
   lzma_index_buffer_decode(&index, &memlimit, NULL, index_ptr, &in_pos, footer_ptr - index_ptr);
   // Just make sure the whole index was decoded, otherwise, we might be
   // dealing with something utterly corrupt
   if (in_pos != stream_flags.backward_size) {
     do_unmap((void *)data, file_size);
     lzma_index_end(index, NULL);
     return -1;
   }
   // Finally get the size
   lzma_vli uSize = lzma_index_uncompressed_size(index);
   lzma_index_end(index, NULL);
   return (size_t) uSize;
}