#1425 Issue closed: RFC: Replace uninterruptible 'sleep ...' by interruptible 'read -t ...'¶
Labels: enhancement, won't fix / can't fix / obsolete
jsmeix opened issue at 2017-07-19 10:26:¶
In current ReaR code there are several 'sleep' calls
with various sleep delay times
(first column shows how often each is used):
11 sleep 1 5 sleep 2 4 sleep 3 1 sleep 5 6 sleep 10 2 sleep 30 1 sleep 60
I think it could be better when instead of an uninterruptible 'sleep'
an interruptible 'read' call is used so that the user could in
particular
interrupt longer sleep delays.
E.g. for longer sleep delays something like this:
delay=10 Print "sleeping $delay seconds, hit any key to interrupt" read -t $delay -n 1 -s 0<&6 && Print "sleep interrupted"
That code sleeps silenty unless ReaR runs in verbose mode.
A possible drawback is that "rear recover" always runs verbose
so that during "rear recover" the user gets unexpected messages
about those sleep delays but on the other hand it is the intent
that the user can interrupt those sleeps.
This issue is somewhat related to the basic ideas behind
https://github.com/rear/rear/issues/1366
and
https://github.com/rear/rear/issues/1399
that both intend to give more power to the user.
This issue here intends to empower the user to
interrupt longer sleep delays.
gdha commented at 2017-07-19 11:02:¶
@jsmeix if the sleep time is less then 5 I would not bother changing the code. OTOH I have objection against the proposal.
schlomo commented at 2017-07-19 12:23:¶
@jsmeix which problem do you want to solve?
What should happen if a user interrupts the sleep? Should ReaR continue like nothing happened or should ReaR abort?
If to change something then I would only improve the exit behavior of ReaR, meaning that ReaR would faster or cleaner abort when the user presses Ctrl-C. Otherwise I assume that all the sleeps in ReaR are there for a very good reason.
If you think that a sleep is not needed then please simply remove it instead of creating a workaround.
jsmeix commented at 2017-07-19 12:30:¶
A plain 'sleep' is basically never there for a good reason.
A plain 'sleep' basically always tells that
the one who made the code had no idea
what the actual thing is that should be waited for
so that he used 'sleep $blind_guess_delay_time'
as some kind of workaround.
I will just leave all as is.
[Export of Github issue for rear/rear.]