| |||||||||||||
|
| |||||||||||||
OLS: Shuttleworth on free software developmentOLS: Shuttleworth on free software developmentPosted Jul 31, 2008 6:18 UTC (Thu) by dlang (✭ supporter ✭, #313)Parent article: OLS: Shuttleworth on free software development
the 'robot trunk maintainer' is a bad idea becouse a robot cannot have any taste. it can tell if the update breaks code that someone thought to write a test for. it has no way to tell if the update does insane things (embedding a flight simulator in a spreadsheet program for example), let alone the much more common 'is this the right thing to do' thinking that the best trunk maintainers do.
(Log in to post comments)
OLS: Shuttleworth on free software development Posted Jul 31, 2008 8:54 UTC (Thu) by james_w (subscriber, #51167) [Link] Hi, Firstly, the robot Mark speaks of already exists, it is called the Patch Queue Manager, or PQM. There is a load of work being done on it right now to make it much better though. Secondly, I work on a project that uses PQM, and I like it. It is no substitute for code review, where taste is exercised, but it does mean that you can have confidence in your trunk. Yes, you can get this with a human gatekeeper, but with thorough code review the merging in to trunk can be done by a machine, so why not let it? (No, PQM can't resolve merge conflicts, so you have to provide a branch that will merge cleanly, usually by merging trunk first)
OLS: Shuttleworth on free software development Posted Aug 1, 2008 5:37 UTC (Fri) by njs (subscriber, #40338) [Link] It's not a replacement for the maintainer, it's a replacement for your vcs's 'commit' command. No vcs that I'm familiar with has a commit command that will stop you from merging silly code to trunk either.
OLS: Shuttleworth on free software development Posted Aug 1, 2008 6:08 UTC (Fri) by dlang (✭ supporter ✭, #313) [Link] with all the good choices of VCS out there, why would you choose to use one that required a replacement for it's commit command instead of just using one that doesn't require such a replacment?
OLS: Shuttleworth on free software development Posted Aug 1, 2008 6:27 UTC (Fri) by njs (subscriber, #40338) [Link] Because none of those VCSes have the functionality you want, i.e., to aid the desired workflow of all development on branches, clean merges to trunk, trunk always passes testsuite...?
|
|
||||||||||||