relblobif should use fcntl instead of flock

[Dimitar Misev]

Currently ​flock is used in source:relblobif to handle transaction locks on the filesystem. The problem, as the man page notes, is that it doesn't work on NFS:

flock() does not lock files over NFS. Use fcntl(2) instead: that does work over NFS, given a sufficiently recent version of Linux and a server which supports locking.

So usage of flock should be replaced with ​fcntl.

[apercov]

After some research, I believe we should leave the implementation as it is now. NFS supports Flock starting with linux version 2.6.12(​https://gavv.github.io/articles/file-locks/). Additionally, after multiple tries and even more research, I couldn't manage to make isLocked() work with fcntl(). Reason for that is that locks provided by the function are not conflicting to each other, therefore we can always substitute them without receiving an error. Conflict is required for F_GETLK(suggested way of checking the lock) to give us return that the file is locked, and we cannot achieve it in context of one process.

[Dimitar Misev]

Ok, I didn't know that flock works fine with such an old kernel version already, it's not indicated in the man pages. So I guess we can close this ticket as invalid.