Change Details

Running stat on a directory gives a size of on ext3 gives a size of 4096. Running stat on a directory on NTFS gives a size of 0. If you copy some files to another FS the total size is different and it falsely appears that some data is lost. We're not trying to count the disk usage, otherwise we'd be looking at block usage not adding up individual files. Therefore this patch removes including the directory size as it's a meaningless value. Note: this patch is separate from "Don't double-count size of directories in DirectorySizeJob". We would still require both. I've split it to separate a potentially controversial design choice from a definite bug.

Running stat on a directory gives a size of on ext3 gives a size of 4096. Running stat on a directory on NTFS gives a size of 0. If you copy some files to another FS the total size is different and it falsely appears that some data is lost. We're not trying to count the disk usage, otherwise we'd be looking at block usage not adding up individual files. Therefore this patch removes including the directory size as it's a meaningless value.

Running stat on a directory gives a size of on ext3 gives a size of 4096. Running stat on a directory on NTFS gives a size of 0. If you copy some files to another FS the total size is different and it falsely appears that some data is lost. We're not trying to count the disk usage, otherwise we'd be looking at block usage not adding up individual files. Therefore this patch removes including the directory size as it's a meaningless value. Note: this patch is separate from "Don't double-count size of directories in DirectorySizeJob". We would still require both. I've split it to separate a potentially controversial design choice from a definite bug.