Don't attempt to write on invalid input

[pramsey]

Not sure what exact input causes this, but it's necessary to regress PostGIS on all platforms

[rouault]

The commit message doesn't match what the code does. What are the situations where p == NULL or nmemb == 0 ?

[pramsey]

Just blindly getting our software running again... postgis/postgis@00eb4b4#diff-7dca6e3da1ca13b272669d5646bcf00bL118

[r-barnes]

This looks reasonable to me.

[thbeu]

I'd claim this should not be merged:

• It's not consistent throughout all I/O hook functions.
• You can define your own hook function for FWrite to fix it on the caller side.