On 2/2/06, Randy W. Sims <[EMAIL PROTECTED]> wrote: > Randy W. Sims wrote: > > Ken Williams wrote: > > >> So, does this mean that someone with this EU::MM::Install would be able to > >> upgrade Cwd on windows now? > ... > >> If the answer is yes, could we get this into EU::MM standalone release > >> too? > > > > Better would be for someone to take over EU::Install which is what > > Schwern wanted to happen, and release it standalone. The current version > > of MakeMaker is already setup for this scenario: > > EU::Install, EU::Installed, & EU::Packlist seem to fit together well > enough, so I packaged them together. All tests pass. > > If this is a good arrangement, we just need the cvs logs for the 3 > modules and their test files. Tests borrow some of MakeMakers > scaffolding; this might could be made more independent if necessary. > > <http://thepierianspring.org/perl/ExtUtils-Install-1.35.tar.gz> > > If Schwern will give the nod, Ken, do you want to add this to the MB > project? Or anyone else interested in maintainership?
Hi, the attached patch cleans up and enhances the original handling of the install at reboot stuff. Its against the seperated out bundle you mention above. It resolves a problem in that the earlier patch doesnt work properly with MakeMaker as the dlls end up readonly and the at boot behaviour fails. This also makes the load of Win32API::File happen only on demand, and adds the delete at reboot ability for things like shadowing. Also there is a minor incompatibility between MB and EUMM with regard to how blib is populated. In EUMM the blib files end up readonly, which EUMM handles by chmodding before copying over. MB doesnt do the chmod first so the copy fails. Thus with EU::Install if you do perl makefile.pl nmake perl build.pl perl buld youll get a failure of copying. you have to nmake clean first and then MB is happy. Im not sure what the best approach is to this so I havent bothered with a patch. IOW both behaviours seem reasonable so its unclear which is really wrong. Unilaterally overwriting files in blib seems reasonable, as does not overwriting readonly files in blib. Cheers, yves -- perl -Mre=debug -e "/just|another|perl|hacker/"
