Cosmetic fix for RHEL 6 and missing name_assign_type for LAN interfac…

[gdha]

Pull Request Details:

• Type: Enhancement

• Impact: Low

• Reference to related issue (URL): cat: /sys/class/net/eth0/name_assign_type: Invalid argument #2197

• How was this pull request tested? Manual

• Brief description of the changes in this pull request: cosmetic fix to prevent the error message "/sys/class/net/eth0/name_assign_type: Invalid argument" - however, the return code is 1 (with the fix, but also with the code untouched). Therefore, it is a cosmetic fix only.

[gdha]

Not merging the PR as it doesn't fix anything

[pcahyna]

@gdha Well, the first part still gets rid of the error message in the log, no?

[jsmeix]

To me that whole code part looks needlessly complicated
and works somewhat indirectly (cf. RFC 1925 item 6a)
and therefore it cannot work really fail safe.

Why testing for existence of a file when reading is intended
i.e. why 'test -f' when 'test -r' is needed?

Why testing "something else" at all and then blindly run a command
instead of just trying the actually wanted command
and checking the return code of the actual command?

By the way:

Why do we use variable names that start with an underscore?

There is no need to redirect STDERR to /dev/null because STDERR
is redirected to the log and the log is there to log in particular errors
so that we can see later by analyzing a log where things start to go wrong.
I see no reason to hide STDERR messages from the log, except exceptions
(e.g. when a command is needlessly oververbose) but then the
exceptional case where an error is hidden needs a comment
that explains the reason behind why STDERR is suppressed.

I think something like

if ! _name_assign_type=$( cat "/sys/class/net/$_netif/name_assign_type" ) ; then
    # On RHEL 6 "name_assign_type" exists but cannot be read
    # see https://github.com/rear/rear/pull/2231#discussion_r322655091
    # therefore we do the following fallback behaviour:
    ...
fi

is more simple and straightforward.

The return code of an assignent of the form
var=$( COMMAND ) is the return code of COMMAND, see
https://stackoverflow.com/questions/20157938/exit-code-of-variable-assignment-to-command-substitution-in-bash
which also shows how to do things correctly for local variables.