The objective of this paper is to determine a set of rules to be done before each release and define the instructions for generating the distribution tarball together with a scheduler.
- We try to release every 1/2 months
- Version numbering (actually we don't follow any rules for this)
- Codenames for releases MUST be funny (until we didn't get a name that can make me laugh, we should not release anything!)
Before any release we have to:
-
Remove warnings
We don't want to fall in the warning nightmare of r1. Releases should contain no warnings with
gcc -Wallor at least no dangerous ones. -
Sync Vala APIs
Keeping the VAPI files the last ones to be developed between release cycles we ensure that we do not have to maintain synced the code with the vapis and it is possible to easily draw the LIBR API evolution by just diffing the vapi directory.
-
Unit test programs
If available, it would be good to have some unit tests to check nothing is broken. Maybe Vala is the way to go when writing tests, because this way we ensure that pkg-config, libr and vapis works in a shot.
-
Test build on different platforms
The same codebase should be compilable on *nix and w32 systems without modifications. It should be also possible to build it with make threads, so using quadcore boxes with
-j8should be a good place for finding race conditions in the build system. -
Remove commented code and review TODO/BUG/XXX comments
While developing a new release, it's pretty common to keep old versions of the code for testing parts of libraries and be able to go back or find bugs while refactoring code or re-doing-it from scratch. This code, should be reviewed and removed if necessary.
$ grep -r -e TODO -e XXX -e FIX libr -
Graph per symbol-module dependency graph to identify unused/dupped/- simplificable use cases of the API for every module.
- Commands should be handled in a structural way, not by a bunch of switch/cases