1. 26
  1. 9

    When I’m searching for something and there’s a link to Julia’s site, I’m always very happy.

    This post about scanimage saved me a couple hours just last week.

    The level of clarity she manages about any topic she chooses to write about is really something special. It’s extra-interesting that she’s exposing this piece as she’s in the process of writing it. I should save this link and look at it next to the eventual final zine to use as an aspirational example for my own writing.

    1. 2

      I’m dealing with one now. My Xlib-based program suddenly stopped working, complaining about a bad window. I’ve tried everything – even undoing all changes since the last working version. No dice. I’m shutting the computer down and going to bed!

      1. 2

        I feel like you’ve pointed at a category of bug similar to but distinct from “distributed system” and “external library you don’t understand” and some of the other ones Julia called out:

        The bug is due to complected systems and their interactions

        The worst problems I’ve ever worked with arise from interactions. Your input tickles a bug in an upstream system, that system’s internal state gets trashed, and then you get timeouts/crashes/wrong answers. Maybe those things expose fragility in the system you care about.

        Then, an indeterminate time later – naturally after you’ve started working on fixing your system, but before you’ve connected the dots – a supervisor daemon/watchdog/coworker reboots/applies a data fix/hits the upstream system with a hammer. Now your carefully reduced testcase starts working again.

        1. 1


          Totally wasn’t that I was too tired to realize that I didn’t really undo the changes because I apparently simply forgot to run git checkout.


      2. 1

        Currently having one of these “Impossible bugs”, so I can relate…

        1. 1

          Dealt with two yesterday….

          • One due to excessive cleverness. Colleague reused a buffer for multiple unrelated purposes and even shared it which worked only due to subtle undocumented reasons.

          • One due to excessive stupidity. Trying to extract common code, only read the first part of a case statement “case FOO_BAH_B…:” and failed to see in one copy it was FOO_BAH_BLAH and the other FOO_BAH_BLUU (case labels names changed to protect the innocent).