Thanks for the response. $pdffile is a path to the "physical" file, and $pdfid is the name of that file, without the extension.

I just tried pulling out this code into a separate script and giving these variables hard-coded values, and it works fine there, so it seems to be caused by something elsewhere than in the snippet I posted. This little PDF subroutine is part of a monstrous pile of code, thousands of lines long, and a real nightmare to debug.

A perhaps related issue is that on some browsers, the download never quite gets wrapped up properly - it gets left with a ".part" extension which needs to be removed manually in order to use an otherwise fine file.

Still digging away at this...