Comments inline. New webrev here: http://cr.opensolaris.org/~error404/6841312_u3/
Vladimir Marek wrote: >>> I haven't seen reply to my question: >>> >>> >>>> First of all, it surprises me a bit that usr/src/cmd/perl/contrib is >>>> not ported. I can not see note about it in the arc case. So perl >>>> 5.10 won't support Sun::Solaris modules? So making >>>> /usr/perl5/5.10.0/bin/perl the default /usr/bin/perl is going to >>>> break customers scripts? >>>> >> I talked to Alan about this and you're probably correct that it should >> be delivered together, so I'll get that done right away. >> > > Ah, I'm sorry to be messenger of band news then. Pity I haven't seen > the ARC case to speak up earlier. I hope that we won't find any > dependency between those modules and ONNV gate ... > I've looked it over and talked to some people and because of some cross dependencies and putback logs (the modules occasionally change when other stuff in ON changes ) I think it makes the most sense to keep the contrib/ modules in ON for the time being, so I guess a phase 1 of this project is delivering perl, phase 2 ( in a build or 3 ) being to update the stuff in O/N to deliver the 5.10 modules ( which I presume needs an ARC case ). So ultimately the modules are coming, just not now, and not through SFW :) >>>> usr/src/cmd/perl510/atinc.patch >>>> >>> I might be missing something, but could you please explain the reason >>> behind changing this? (it was not present in previous version) It >>> seems >>> to set some paths for tainted mode, and the paths are >>> /usr/local/lib/perl5:/usr/local/lib/perl, which might not be what we >>> want? >>> >> it places site_perl and vendor_perl at the beginning of @INC, the patch >> was suggested by Mark Trettin and it makes sense to me >> > > I see. To me it would make sense, if customer could override perl > modules by placing them into site_perl. Hopefully it won't break > anything. I just wonder that the patch says '/usr/local/lib/...' > > 35 +#ifndef PRIVLIB_EXP > 36 +# define PRIVLIB_EXP "/usr/local/lib/perl5:/usr/local/lib/perl" > 37 +#endif > > We don't have anything in /usr/local/lib, but maybe PRIVLIB_EXP is > defined earlier, so it wont' pick those paths. > It's defined earlier, That whole file is scattered with references to /usr/local. Vestige of GNU I guess. They don't affect the final outcome > However, please, once you'll be testing new perl, try that the @INC is > sane for normal and tainted mode. Something like: > Testing is assured. @INC is normal ( for posterity here: ) $ ./perl-ws/proto/root_sparc/usr/perl5/5.10.0/bin/perl -T -e 'print join "\n", @INC' /usr/perl5/site_perl/5.10.0/sun4-solaris-64int /usr/perl5/site_perl/5.10.0 /usr/perl5/vendor_perl/5.10.0/sun4-solaris-64int /usr/perl5/vendor_perl/5.10.0 /usr/perl5/vendor_perl /usr/perl5/5.10.0/lib/sun4-solaris-64int /usr/perl5/5.10.0/lib > I'm quite sure you already hate me, but have you seen my comments about > config.sh? I believe that few configuration options are wrong there... > Yes, this new webrev brings config.sh(.*) much more in line with the one in ON (sorry I thought I had responded to that point but I guess I didn't ) Thanks a lot -JohnS
