This is my busiest month of the year for my day job, but eventually I want
to look into the date code in POI, as I suspect there are some parts
handling this case explicitly and perhaps this path and a few others that
could benefit from some standardization around whatever that behavior is.
Yeah, I feel like I should be able to create a simple example that doesn't
use the conditional stuff so far but haven't been able to change it much
without, uh, brekaing the brokenness. So I think there's something catching
the issue in most cases.
Sounds like an input validation issue with EDate and other similar
functions to me. With Conditional Formatting I tried to not swallow/ignore
errors in evaluation, exactly because I didn't want to hide bugs like this
appears to be.
Excel seems to assume a double of -1 equals a date of 0/0/1900,
