This sort of problem is frequently caused by trying to pull from a repository that has two filenames that differ only in case. If you are on FAT, NTFS in case-insensitive mode (essentially, any time it's being used under Windows), or HFS+ in case-insensitive mode, and have two files "foobar" and "FOOBAR", then Git will see two distinct files, but the filesystem will only see one, which will cause all kinds of problems. Git will checkout, say, "FOOBAR", and then checkout "foobar", which the filesystem sees as simply replacing the contents of "FOOBAR" but leaving it in place. Now to Git, it appears that "FOOBAR" has been replaced with the contents of "foobar", and "foobar" is gone.
There are two different manifestations of this basic problem. One is when your repository actually contains two files that differ only on case. In this case, you need to work on a case-sensitive file system, or you will need to edit the repository to ensure that no collisions of this sort occur; a case-insensitive file system simply cannot store the contents of this repository.
A different case that you can workaround is when a rename happens that changes the case of the file. Say, for example, that the Git repository contains a rename from "EXAMPLE" to "example". Before Git checks out the new version, it will try and check to make sure it's not overwriting some existing file that you have on your disk. Since it thinks that "example" is a new filename, it will ask the filesystem if it exists, and the filesystem will see "EXAMPLE" and say yes, so Git will refuse to check out the new version since it thinks it will be overwriting untracked files. In this case, if you have no local changes that you care about, a simple git reset --hard <revision-to-checkout>
will generally be sufficient to get you past the problem and to the new revision. Just try and remember not to rename files to other names that differ only in case if you're on a case-insensitive file system, as it will cause problems like this.