Erik and deja_vu - next time you see this, please do the following and report your findings:
1.) Run `stat` on the decrypted file and paste the results.
2.) Run `stat` on the encrypted file and paste the results.
3.) Run `hexdump -Cn 8` against the encrypted file and paste the results.
4.) Remount the eCryptfs mount.
5.) Repeat steps 1 through 3.
Unfortunately, if you're using encrypted filenames, it is going to be difficult to figure out the encrypted filenames in order to run stat and hexdump against those files.
Erik and deja_vu - next time you see this, please do the following and report your findings:
1.) Run `stat` on the decrypted file and paste the results.
2.) Run `stat` on the encrypted file and paste the results.
3.) Run `hexdump -Cn 8` against the encrypted file and paste the results.
4.) Remount the eCryptfs mount.
5.) Repeat steps 1 through 3.
Unfortunately, if you're using encrypted filenames, it is going to be difficult to figure out the encrypted filenames in order to run stat and hexdump against those files.
Also, if you don't mind running a bleeding edge kernel, I rewrote the previously buggy truncate path and that patch (http:// git.kernel. org/?p= linux/kernel/ git/torvalds/ linux-2. 6.git;a= commit; h=5f3ef64f4da1c 587cdcfaaac7231 1225b7df094c) was released in 2.6.33-rc5. If I had to put my money on it, I'd guess that this is happening after truncating the eCryptfs inode.
I ran the git test suite in an eCryptfs mount several times and it only complained about the same tests that failed on plain ext4.