(For some reason Slack web version didn't show me these replies. I was wondering why nobody else was thinking about this.)
The reason I was thinking about this yet again was that I was about to apply the most recent java and cf updates, and wondering what due diligence looks like for various devs and organizations. Not at all ironically, we got bit by
this bouncycastle error, because we didn't test PDF generation. Clearly we should have, we've seen other java updates break it too.
Side notes:
• That issue and the download for the fix should have been listed on the CF downloads page.
• That issue is marked as deferred, because it's a "Duplicate of
CF-4215479" (not the usual meaning of "deferred"!), but that other issue can't be found, apparently.
Anyway, this is quite a big app, written over many years by many people, legacy in many ways though not as spaghetti-d as some. The vast majority of it at least has methods to get the data that are separate from the code to render it, and those back-end methods should be more or less testable, with some work.
Unmanned testing of front end stuff is much harder IMO, and that's where the PDF error would have shown up. But at least it's a crash, easier to trap for than incorrect HTML or broken interactions.
We'll see...