Hi, > How does defaulting to master fix this issue? Isn't it user error?
Yep it is user error but doing so would help minimise those sort of errors. > Our documentation on the website very clearly shows how to checkout from > the develop branch. As does the readme and several other places but I guess people don't always read or understand the instructions. > This is a problem with how Apache Infra is hooking into the external tool. With github that's correct but other places have had similar issues (eg ohloh.net) > Better messaging? Perhaps users/people outside the project may not know the details and it's easy fpr them to make wrong assumptions. > Did mentors advice us to use master branch as default? Please point to the > thread? http://markmail.org/message/wxnrrtcvi2bmgcdv It's re SVN and working in trunk which is equivalent. See also the Git vote discussion thread(s). >> 6. Not in line with other Apache projects. > > Examples? Some examples include: https://github.com/apache/cordova-ios https://github.com/apache/lucene-solr/tree/trunk https://github.com/apache/couchdb https://github.com/apache/subversion/tree/trunk (although it also have a large number of branches) https://github.com/apache/tomcat https://github.com/apache/httpd/tree/trunk With with a few it a little hard to tell but I've not been able to find one into one that doesn't update trunk/master frequently. > On the contrary. Having a develop branch ensures that bad check-in does > not pollute the main codebase. A check can be easily reverted/rolled back and with Git's easy branching it should be easier to keep untested/unstable code out of master. If we worked in master there would be far less work/time spent merging or no need or merge at all in a lot of cases. Makes making releases easier. Users want 1) release builds 2) latest stable build 3) latest build, currently we can only do 1 and 3 and it's a long time between releases. Releasing more frequency would also be an answer but people don't seem to be interested in doing that as it's a time consuming/complex process. > With what you are suggesting, master will be as unstable as the develop > branch is today. I agree only stable code should be placed into master, we should just try and do it more frequently than after each release. Currently master while "stable" (it's at the 4.9.1 release mark) is missing a lot of fixes so it could be argued that it's actually less stable/more buggy than develop. We basically have 2 branches (master and release branch) that have exactly the same info in them - seems redundant to me. With the CI up and running we should know (most of the time) fairly quickly when there is an issue. Thanks, Justin
