[jbosstools-dev] Re: State of SVN and build

Max Rydahl Andersen max.andersen at redhat.com
Tue Apr 29 00:24:25 EDT 2008


>>> Nothing prevented you from doing a component specific tag/branch for
>>> your experimental work.
>>>
>> Except the idea that I shouldn't ever have to work in a *branch* to do
>> forward dev on my own components that *I* own. Obviously if it's a
>> shared component, the way xulrunner was introduced, a branch is
>> necessary. But when I'm the owner of a component, I should be able to
>> push my forward-progress (or experimental as you call it) in HEAD.
>>
> And there's nothing stopping you from building your
> Ganymede-bleeding-edge code in trunk. Just let us know before you do so
> we can point the build at the proper "3.3" branch so nightlies work as
> expected.

Marshall - if we do that trunk becomes inconsistent and we loose the current *valuable*
setup where we can just checkout trunk.

>> But of course this also means that if someone does check out trunk,
>> there is no guarantee it will all compile together as each of the
>> trunk portions of the component could be targeting different drivers.
>> So to get a successful build, they'd have to check out all from trunk
>> and then replace the ones that have moved forward with their branches
>> that stopped at current dev.
>>
>> Obviously this is just as complicated and what you're trying to avoid.
>>
>> I can be swayed to keep my HEAD at current jbds driver levels and do
>> my future dev in a branch, SO LONG AS the branches folder is neat and
>> organized and I have my own component folder there
>> (branches/components/as/futureDevGanymede) for example.  This way I
>> don't have to worry about coliding with other obscure names that may
>> follow no apparent convention =P
> Releng doesn't care if bleeding-edge dev goes in trunk or a new branch,
> it's all the same. That being said, we offer nightly builds for a
> reason... if a potential contributor wants to look at the latest and
> greatest for a specific version, maybe we should start providing PSFs
> again to automate the code checkout process.

Why make things complex for the common case ? 

If a component want's to do things that does not fit into jbosstools release
cycle he branches.

The day every/most component are doing this then lets come back to it.
Right now the current setup it is the simplest that work.

/max






More information about the jbosstools-dev mailing list