create_directories: preserve sticky and setgid permissions

[franzpoeschel]

Ref. ComputationalRadiationPhysics/picongpu#5638

[franzpoeschel]

Why do we retrieve the umask just to apply it manually without any modifications? That's what the system does implicitly anyway?

[franzpoeschel]

Is that supposed to make the function thread-safe against modification of the umask on other threads? If yes, this does not work due to the double call to umask(), also it is irrelevant for MPI-parallel deletion.

[ax3l]

Agreed. Let's make sure that if someone creates dirs and files with openPMD, it behaves the same as a terminal mkdir and touch with respect to umask and other inherited properties.

It's possible the current implementation is overly verbose and forgets setgid/sticky bits

[franzpoeschel]

Discussed yesterday: Our current implementation probably assumes that the umask is overwritten by the mode_t mode parameter of mkdir(), but that is not the case:

       The  argument mode specifies the mode for the new directory (see inode(7)).  It is modified by the process's umask in the usual way: in the absence
       of a default ACL, the mode of the created directory is (mode & ~umask & 0777).  Whether other mode bits are honored for the created  directory  de-
       pends on the operating system.  For Linux, see VERSIONS below.

Hence, our manual umask handling (1) is unnecessary and (2) is not thread safe due to temporarily manipulating the umask.