On 12/30/19 2:32 PM, Daniel Wagner wrote:
Indeed, the error handling is very difficult. I was only looking at
patch itself without considering the big picture. So the thing I was
pondering on was following:
[ a: "foo", b: "bar", UserRoutes: [...] ]
'a' and 'b' will be parsed and stored via vpn_provider_set_string(). If
there is an error in UserRoutes parsing, 'a' and 'b' will not set the
the old value (reset).
As I understand the error will be reported via an D-Bus error message
and there is no disk write at this point for 'a' and 'b'.
Not sure if we enter the over engineering area here :) You fixed your
initial problem (reducing the disk writes) and the error handling
is still as good it was. Let's see how this works and if we see real
world problems we can still address them.
Mondays after holidays usually induce the over engineering part. I
agree, lets leave this part untouched for now.