On Thu, 9 Sep 2010, Gustavo Sverzut Barbieri wrote:
> Hi all, > > After lots of people complaining to me at IRC, mail and others, I'd > like to ask you developers to improve commits and particularly their > messages. > > 1 - The most complained annoyance is by far commits that break SVN, > making it unreliable for users... and if you recall we don't have > releases to point to users and have them to stay away from live svn > ;-) Okay, people will say (me included) SVN doesn't make our lives > easy and $ANOTHER would help, but raster already said "SVN IT IS", > thus let's try: > > - svn status -- without -q, check lines starting with "?" and > either "svn add" them, or "svn propedit svn:ignore $directory" and > list it there. This way we avoid missing files (and missing ignores!) > > - svn diff, read what you're committing, see if it does make > sense, if you're committing more than you should (tests, debug, etc) > > - make && install && test. If you added a compile-time toggle, > test with and without your option > > > 2 - The second most complained is commit messages. Okay, we're a fun > project and we all commit funny messages, but try to provide some > insightful details, try to make the first line 72 chars and make it a > summary. Please comment the design of your newly added feature, if > there are missing details comment it too. If it is a bug fix, then try > to explain how you fixed it, maybe a test case that verifies your fix. > If it is a leak fix, some details on how you found it, how you fixed > it. -- I do know some cases are harder to provide tests, like E17, > but then let's try to be more descriptive! > > 3 - The third most complained is mixed commits. Particularly bugfixes > that contain formatting and whitespace changes. Please either commit > the cleanup first, then the fix or vice versa, but not together. "svn > diff" may help you there. These mixed commits are nasty because often > the bug fix is not complete, or breaks something else, and when you > look at the diff you immediately say "WTF, 500 lines of changes to fix > this?!!?!? It will take lots of time to figure out the changes". And > usually the formatting/whitespace is tricky, because lines look alike, > then you start to ignore or waste hours staring at a line to figure it > out. Consider the following example (not real): > + > - > + fnc(); > - fnc(); > + x = i + l; > - x = j + 1; > + // comment heer > + // comment here > Now make it hundreds of blocks like this, it is quite hard to spot > that "x = i + l" changed to "x = j + 1", as you start to ignore stuff > as they are all irrelevant in the sibling lines. > > 4 - The fourth annoyance is related to the previous and is could be > called "commit torrent" or "ssh over svn" or "gcc over svn" and is the > result of people committing every line or test they do, then > committing couple of changes, then commit removing debug code, then > another fix... I don't know the reason, maybe people want to copy > stuff to their compile servers, and instead of scp it there, they > commit and the compile server automatically svn up && make, but it is > annoying. These commit should all be merged/squashed into a single one > and avoid people running into them. > > > If you follow SVN you'll definitely know the dude that tops all the > annoyances is our beloved top committer: Rasterman :-D As I complained > a lot to him about these facts in IRC, I know of some excuses, some I > do agree like "increase number of testers", but he could do better. > Problem is that he is the "example to be followed" and thus we tend to > get worse and not better, thus I again would like to see some > consideration here. > > As we're approaching a release, I'd like to ask you all, including > Raster, to try to be kind and improve over these topics. > > Comments? Suggestions? using the moap tool Vincent ------------------------------------------------------------------------------ This SF.net Dev2Dev email is sponsored by: Show off your parallel programming skills. Enter the Intel(R) Threading Challenge 2010. http://p.sf.net/sfu/intel-thread-sfd _______________________________________________ enlightenment-devel mailing list enlightenment-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-devel
