About change logs

Versions of Inform 7 are identified by "build number", a combination of digit, letter, and two digits: for instance, 2P75 comes after 2P74 and long after 2G14, but before 2Q02 and long before 3A07. Inform story files normally print some versioning text in the "banner" fairly early on in play, and often immediately. For example:

Release 12 / Serial number 090410 / Inform 7 build 5Z71 (I6/v6.31 lib 6/12N)

Each public build has its own change log, which notes only what has changed since the previous one. Minor changes to documentation are not listed.

Every public build of Inform 7

In reverse chronological order, that is, latest first:

Build numbers in between these were private ones, and you'll never see a story file compiled by them (unless it was published by one of our beta-testers), with two exceptions:

• 4K40 was rapidly followed by 4K41, identical except that it fixed a glitch on Macs with PowerPC processors. (It was early in Apple's transition to an Intel-based architecture, and we were feeling our way.)
• 4U65 was similarly followed by 4U67 for Mac OS X only, though that time it was to improve the user interface application; the underlying software was the same. (We probably shouldn't have updated the build number.)

Builds 1A01 to 3K26 did exist. Inform 7 was nearly three years in the writing before it was ready to be offered as a public beta. 1A01 was the first even slightly working version. It did not synchronise fully with the OS X Inform application, the first to be written, until 1G22. Private beta-testing did not begin until 1J34. Other milestones include time (1B92), tables (1C86), component parts (1E60), indexing (1F46), systematic memory allocation (1J53), pattern matching (1M11), the map index (1P97), extension documentation support (1S39) and activities (1T89). The first round of testing came informally to an end at around the 1V50 build, after which a general rewriting exercise began. Minor changes needed for David Kinder's Windows interface began to be made with 1W80, but the main aims were to increase speed and to improve clarity of source code. Hashing algorithms adapted to word-based syntax began to be introduced in 1Z50; by 2D52 there was a speed increase of a factor of four. A second stage of rewriting, to generalise binary predicates and improve grammatical accuracy, began with 2D70. There was then an exhaustive round of testing, putting together the first sets of test cases, before 3K27 was published - though, as the rapid succession of updates testifies, it was still fragile enough. The first decently robust build came a year later, with 4S08.