Michel Juillard michel.juillard@mjui.fr writes:
I think that having a GIT branch identical to the snapshot would be useful when investigating problems reported in the snapshot, particularly if master becomes less stable than now.
Ok, then we keep the separate branch.
Currently, the unstable version is publicly known and the name is used in the Download section of the web site. I have been telling people to download the "unstable" version for several months. If we change the names, we will need to be very explicit on the web site in order to avoid useless queries.
What we can do is to distribute two snapshots: the "unstable" (corresponding to the master branch) and the "testing" one. Thus we are not contradicting any of our previous statements: we have always said that the "unstable" snapshot can be broken. We can present "testing" as a new snapshot which is between "stable" and "unstable" in the stability/bleeding-edgeness space.
Of course most of the time, today's testing will be equal to yesterday's unstable, but the advantage is that this setup allows for a temporary divergence between the two branches (which could be useful if for example we want "testing" to be temporarily equal to a broken "master" + some dirty workaround that we don't want to incorporate into master).
On 09/19/2011 02:08 PM, Stéphane Adjemian wrote:
I agree with Sébastien. A "testing"-snapshot version would be enough if we distribute the windows binaries *and* the sources... Because the idea is not to commit directly in the "testing" branch but to automatically cherry-pick (in testing) commits from master. Best, Stéphane. Sébastien Villemot <sebastien.villemot@ens.fr> writes: Michel Juillard <michel.juillard@mjui.fr> writes: It is very urgent to put in place the testing version that we talked about. I have referred several people to the unstable version for things that I fixed over the Summer and we can't having it broken over long periods of time (more than one day, I would say) In fact, intstead of having stable, (new) testing, unstable, I would suggest that we have stable, unstable (that we try to keep in workable state), (new) experimental which is not supposed to be entirely working. The names will be a bit misleading but it will avoid to have to inform people who download unstable to have to download testing instead. Personally I prefer the name "testing", it corresponds more to the reality. However, I realize that if we want to implement this "testing" version only for users who download the snapshot (and not for those who use git), there is no need to create a new branch. I could just modify my snapshot building script so that a new snapshot is only created if all tests pass in the testsuite. The lastest snapshot available for download would therefore correspond to the last working state of the git. From the point of view of users who download the snapshot, this is "observationally equivalent" to having a testing branch. _______________________________________________ Dev mailing list Dev@dynare.org https://www.dynare.org/cgi-bin/mailman/listinfo/dev
Dev mailing list Dev@dynare.org https://www.dynare.org/cgi-bin/mailman/listinfo/dev