On Tue, Aug 13, 2013 at 12:31:12PM -0400, Eric Whitney wrote:

Clearly, I wasn't thinking about the ABI at all - thanks for pointing out
that misstep.

So, I'd like to withdraw that patch, please, and will post a V2 in a bit.
Computing the eof_blk in that manner is better than an initial patch I had
that worked but which was pretty ugly.

It didn't really make sense to put the eof_blk in the extent_info
structure, anyway, since it's not information about the that specific
extent.  It's an inode-specific value which is set via:

	handle->path[0].end_blk =
		(EXT2_I_SIZE(handle->inode) + fs->blocksize - 1) >>
		 EXT2_BLOCK_SIZE_BITS(fs->super);

Right?  So this is something you can calculate without making any
changes in lib/ext2fs/extent.c, unless I'm missing something.

							- Ted