[postgis-devel] 2.2.2 Release (?)
gdt at ir.bbn.com
Tue Mar 1 06:50:52 PST 2016
Sandro Santilli <strk at keybit.net> writes:
> On Mon, Feb 29, 2016 at 08:48:48AM -0500, Greg Troxel wrote:
>> My impression is that separating this wouldn't be that hard. The big
>> issue is separating "checkout as modified but not committed" from
>> "what's in the working dir". But even if it could easily make a
>> distfile from trunk or a branch, vs a tag, that would help a lot.
> The only reason why that script relies on the VCS is to avoid putting
> in the tarball any generated file, basically.
> Automake solves that by requiring explicit mention of each and every
> file that needs to be in the distribution (in Makefile.am).
> Do you think we could use a manually-edited MANIFEST file for that ?
We could, or we could use svn ls on the branch. Once we have a
manifest, I'd want to have 'make distcheck', but this is slowly
reinventing automake which isn't a good strategy....
But in thinking about this, the real benefit is to be able to make a
distfile from trunk, or the head of some branch, to check it during the
runup to a release. So if the script could make a distfile from trunk
or a branch, with read-only VCS access, that would get 85% of the value
with perhaps only 10% of the work.
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 180 bytes
Desc: not available
More information about the postgis-devel