Zero deemed invalid lower bound for interval censored data in survreg with lognormal

[chjackson]

survreg produces "Invalid survival times for this distribution", when interval-censored data for the lognormal have a lower bound of zero.

emin <- c(0, 1:10)
emax <- emin + 1
survreg(Surv(emin, emax, type="interval2") ~ 1, dist="lognormal")

I agree zero is invalid for a log-normally distributed event time, but I think it should be valid as a lower bound with interval censoring, or have I missed something? This also happens for the log logistic - I haven't tested for others.