[SciPy-dev] Trac-ing backports

David Cournapeau david at ar.media.kyoto-u.ac.jp
Thu Jun 4 20:48:03 EDT 2009


Pauli Virtanen wrote:
> Hi,
>
> I noticed that keeping track of bugs to backport is a bit painful. Also 
> Trac isn't currently too helpful about this.
>   

No, it is not :)

> So, how should we deal with backports within Trac:
>
> a) For each bug you think could be useful to backport, before closing the
>    ticket, file a new "Backport #xxx to 0.y.z" ticket. Easy to forget,
>    though, and a bit nasty with Numpy with its 1.0.x, 1.1.x, 1.2.x, 1.3.x,
>    branches...
>
>   

That's the recommended approach (http://trac.edgewall.org/ticket/4298),
but I find it awkward. It means a same bug has different number
depending on the milestone.

> b) Add a custom field "Branches" that contains some custom format, such as
>    "-1.3.x" before and "+1.3.x" after backporting. Add corresponding
>    queries to Trac's roadmap.
>   

So every backport would require a manual change to trac as well ?

What about having a special format for backport in the svn commit
message, so that it can easily be grepped ? Something like
BACKPORT:ISSUE#NNN as the first characters of the commit ? At least we
could track this from git log --grep this way. I am afraid that
requiring trac handling with backport will result in a lot of misses.

cheers,

David




More information about the SciPy-Dev mailing list