Some time ago, there was a performance-related bug that went something like this:

mm/dd/yy Created by bob
The attached file contains a dataset that takes a very long time to process.

The engineer who fixed the problem decided to take the cryptic approach:

mm/dd/yy Resolved as fixed by alice
It got better.

It was a common practice during highly stressful periods to file humorous bugs in the defect tracking system, and once the initial bug was filed, it turned into a sort of collaborative performance effort. For some reason, milk is often a trigger. Many years ago, the vendor who supplies milk to Microsoft changed the design of their milk cartons, and the new cartons were significantly harder to open than the old cartons. (Apparently they bought a new machine, and the glue level needed to be tuned.) This naturally inspired people to do what they usually do when they see something amiss: File a bug. More than one person was inspired to file a milk carton bug, but probably the most famous one is the one reprinted by Korby Parnell back in 2003. Kraig Brockschmidt describes an earlier milk carton bug:

A friend of mine who worked as a tester on Microsoft Excel once logged a bug against the cartons of milk in the free drink coolers. He noticed that the chocolate milk failed to list the ingredient “cocoa” whereas the plain 2% milk did. An intense discussion in raid [the defect tracking system] over the relative merits of these “features” continued for three or four weeks and involved as many as a fifteen different engineers. I think it set some kind of record. Anyway, someone finally went so far as to notify the dairy itself and the bug was closed as “won’t fix—assigned to vendor.”

(In the years since the original story occurred, our defect tracking system gained a new resolution class, and today it would be resolved as “external—assigned to vendor.”) Over in the languages group, there was a bug that went roughly “IDE fails to beep when build is finished.” Nobody can find a copy of the bug any more, it having been archived away ages ago, but one of my colleagues recalls that it went something like this:

• Start a build.
• Leave office to get a beverage from the kitchen.
• Return to office.
• Observe that build has finished, but no beep was heard.

One person resolved the bug “Not Repro” because they stood just outside the office and clearly heard the beep, but that led to a debate over how far from the office you needed to be in order to reproduce the bug. Things got sillier and sillier until someone finally closed the bug after claiming to have taken their computer out to the forest behind the building and started the compilation. When they returned 30 minutes later, they found that a tree had fallen on the machine.

I admire the philosophical resolution to that one.

Bonus content: Mark Lambert posted the text of the “IDE fails to beep” bug.