Pádraig Brady wrote:
A quick scan shows that functions using tz handle the NULL case, falling back to default behavior.
Yes, when I wrote that code I assumed that users prefer getting UTC time stamps to getting a diagnostic and no time stamps. That's the Unix tradition for other tz setting failures, e.g., TZ='no-such-time-zone'. If my assumption is wrong and users prefer an error message when memory is low, the patch would make sense. It's low priority, though, as the behavior is well-defined whether or not the patch is installed.
