Is it always encouraged to upgrade spkgs to more current stable
upstream version?
Right now scipy is at 0.9, applied with patches to fix a bug that is
now corrected in version 0.11 (normally, I have not tested yet). The
patch does not seem to work under Mac OS X 10.8, so I was wondering if
instead of fixing the patch I should really upgrade the spkg to the
later version for everyone.
Paul
On 16/10/2012, at 7:17 PM, Paul-Olivier Dehaye <paul-olivier.deh...@math.uzh.ch> wrote:
> Is it always encouraged to upgrade spkgs to more current stable
> upstream version?
> Right now scipy is at 0.9, applied with patches to fix a bug that is
> now corrected in version 0.11 (normally, I have not tested yet). The
> patch does not seem to work under Mac OS X 10.8, so I was wondering if
> instead of fixing the patch I should really upgrade the spkg to the
> later version for everyone.
> Paul
> Is it always encouraged to upgrade spkgs to more current stable
> upstream version?
No. I'd say it's neither encouraged nor discouraged.
Usually patching a package is easier than updating to the latest
upstream version. Sometimes upstream upgrades break stuff. But if it's
easy to upgrade, then sure, why not do it?
On Tuesday, October 16, 2012 8:19:11 AM UTC+2, yomcat wrote:
> On 16/10/2012, at 7:17 PM, Paul-Olivier Dehaye <paul-oliv...@math.uzh.ch<javascript:>> > wrote: > > Is it always encouraged to upgrade spkgs to more current stable > > upstream version? > > Right now scipy is at 0.9, applied with patches to fix a bug that is > > now corrected in version 0.11 (normally, I have not tested yet). The > > patch does not seem to work under Mac OS X 10.8, so I was wondering if > > instead of fixing the patch I should really upgrade the spkg to the > > later version for everyone. > > Paul
Given the choice between update (drop a patch that is fixed upstream) and fork (maintain a separate patch), I think we should always opt for the former. Otherwise we'll just get bogged down in an ever-increasing maintenance headache. Don't reinvent the wheel.
So I'd say we should always upgrade to new upstream stable releases. Sometimes that uncovers bugs, which we then can report upstream.
> Given the choice between update (drop a patch that is fixed upstream)
> and fork (maintain a separate patch), I think we should always opt for
> the former. Otherwise we'll just get bogged down in an ever-increasing
> maintenance headache. Don't reinvent the wheel.
> So I'd say we should always upgrade to new upstream stable releases.
> Sometimes that uncovers bugs, which we then can report upstream.
It depends.
For example, in "debian stable", the adjective refers to the fact that the bugs are stable, precisely because things aren't updated to latest upstreams. The downside is that it is always using several years old software, even when it gets out! And that makes sense...
On the other end, when you're looking for say an elliptic curve algorithm or trying to make some things explicit in some very specialized domain, then you are generally interested in the latest and greatest.
