survmean: uninformative error

[WetRobot]

When survival curve estimation fails and consequently a test fails within survmean, this is what the user sees:

The total person-time was zero in some survival intervals, when summed to the variable(s) 'sexaGwHW68X5m', 'agegrsBcfXLlUQx' (i.e. over all other variables, if any) _non-consecutively_, i.e. some intervals after an empty interval had person-time in them. Keeping all survival intervals with some estimates as NA for inspection.
Some cumulative surv.obs were zero or NA:
   sexaGwHW68X5m agegrsBcfXLlUQx first.bad.surv.int last.bad.surv.int surv.obs
1:             0          [0,50)                 34               108      NaN
2:             0         [50,60)                 15               108      NaN
3:             0         [60,70)                 68               108      NaN
4:             0        [70,Inf)                 44               108      NaN
5:             1         [50,60)                 23               108      NaN
6:             1         [60,70)                 22               108      NaN
7:             1        [70,Inf)                 75               108      NaN
Error in if (any(mi$surv > 1)) { : missing value where TRUE/FALSE needed

At the very least if (any(mi$surv > 1)) should be robust for NA values. Much more preferably
survmean should either fail when there are NA values (with human-readable error msg) or automatically try to adapt to the situation (remove all unusable estimates, shift extrapolation to start earlier).