On Mon, 14 Dec 2009 15:06 -0800, "Tim Rice" wrote: > On Mon, 14 Dec 2009, Charles Wilson wrote: > > > Two related lines of inquiry: > > 1) Under *normal* development rules -- e.g not a pre-release > > bug-fix-only phase, nor a not-quite-pre-release code slush like (I > > think) we're in right now, for 2.2.8 -- surely you aren't suggesting > > that EVERY contribution must be validated on EVERY platform, prior to > > push? These were tested on cyg/ming and linux, so in general, during > > /normal/ development, that should be sufficient contra reveiwer > > comments, right? > > I would strongly advocate testing at least one System V based system. > Non bash shells sometimes have issues.
During the release process, certainly testing on as many supported platforms as possible is baked into the cake. *Perhaps*, by a reviewer or other list denizen, before a particular patch is pushed. But it's unreasonable to *require* every contributor to have access to multiple platforms. Some folks who fix HP-UX might not have (or want to install) cygwin on a win32 box. Or a mingw developer might not have any access to a Solaris machine. I have access to cygwin, mingw, and linux. I don't have access to SysV, and it would be setting WAY too high a bar for contributions to say "No, thanks, we don't want your patches because you can't test them on System Foo". -- Chuck
