a1246e41a4
To do this, we have to define the mapping of change -> commit. In the default case, we just pick the SHA-1 of the latest patch set. In a more complex case, like listing related changes, we might be interested in only a subset of patch sets, in which case we pick the SHA-1 of the latest patch set in that subset. Once we collect the mapping of change to commit, we can mark each commit interesting, and do a simple RevWalk until we've seen every interesting commit. This produces the order of changes within a single project. We also have to deal with multiple projects, for example if we want to sort all changes across all projects matching a particular topic. To do this, we keep groups by project together, and sort each group based on the name of the project. This is more stable than alternatives involving commit or change update time, since those are more likely to change while a user is navigating among a list of changes in the project. Change-Id: I9a1fc7dbf308d1e4bde4d0ff1acf10b8aacb9b0f