|
| From: | Richard Frith-Macdonald |
| Subject: | Re: GNUstep win32 release policy |
| Date: | Mon, 13 Mar 2006 07:19:00 +0000 |
On 13 Mar 2006, at 04:28, Alex Perez wrote:
Jeremy Bettis wrote:Alex Perez wrote:Shouldn't we make it a formal policy to test under mingw before making any RELEASE? Is there a checklist everyone much follow before making a release? It should probably be on the wiki, perhaps locked/in review-submit-only mode so it can only be edited by the respective package maintainers.You can't have every platform tested for every release. Even for GCC, windows is on the non-critical list. I am a heavy Mingw user (the only heavy mingw user of gnustep?) and I don't care if Mingw is tested before every release. Frankly the idea of someone who never normally uses windows dual booting over to windows once every 3 months to test under mingw before calling the version RELEASE, doesn't really boost my confidence any.What do you think of this suggestion?That's not really the point. I think both RFM and Adam, and Gregory can dual-boot, and if a given release doesn't work under win32, that's FINE, as long as it's known/stated (and not buried in a release file somewhere, on the website download page where people will actually see it) It's not hard or too cumbersome to compile - base and -gui under win32, and perhaps run the test apps/unit tests on that platform before we release. If we don't do this, the win32 side of things will NEVER get any better, since people won't easily be able to contribute code/test GNUstep under win32.
For what it's worth, I agree that we should test win32 before each release ... but I don't know that Adam and Gregory have win32 setups ... I do right now, but I didn't at the last release (actually, what I had was a completely dead intel workstation and a faulty ppc laptop at the time). It's quite likely they don't run windows routinely, in which case getting everything up to date on their systems might involve checking library versions and downloading stuff. Last time I had to do that I was frustrated by websites being really slow or down, so I know it can take hours. Bottom line is that testing on several platforms is probably just too time consuming when making a minor/snapshot release. Major releases are another matter ... we should have branches for them and be testing them over a matter of weeks, rather than trying to fit time to make a release into an odd hour or two. In those circumstances plenty of people ought to be able to test on different platforms.
It would be nice if someone who actually uses windows all the time could add their machine to Adam's testfarm ... so we would have a nightly check that things at least compile. I don't know how much the testfarm does in terms of runtime checking, but even if it's none at all (and more checks could always be added) the nightly compilation report is very useful.
Actually, a possible enhancement to the testfarm would be if it was aware of release branches somehow, and could test the latest release branches as well as the trunk.
| [Prev in Thread] | Current Thread | [Next in Thread] |