patientlywaiting

When programs are simple, bugs are relatively easy to find and squash. But by all accounts, the TextBlade firmware is amazingly complex (which it would have to be in order to support its revolutionary capabilities). Adding a feature to code_block_A seems simple enough and could be perfectly fine on its own (like the BASIC word-jumble upgrade example). But then code_block_B expects code_block_A to work more narrowly and that causes a bug.

But more realistically, in a project this size, code_block_B is fine with it, but code_block_C depends on B, and D depends on C and...Z depends on Y, and it is code_block_Z that expects code_block_A to work the old way and throws a bug.

And then 1 out of a hundred testers trips over this bug, but it is nearly impossible to reproduce. But WayTools cares very much about this level of quality, so they can't just let it slide. And they can't just revert the code, because now code_block_J and code_block_Q rely on code_block_A working the *new* way. So it takes weeks to find a way to reliably reproduce the bug, and more time to fix it, and then more time to test it, which uncovers more edge cases, which requires weeks to reliably reproduce the bug, which requires...