On Mon, Oct 08, 2007 at 02:24:08PM -0500, Chris Lawrence wrote: > On 10/8/07, Dirk Eddelbuettel <[EMAIL PROTECTED]> wrote: > > On Mon, Oct 08, 2007 at 02:08:54PM -0500, Chris Lawrence wrote: > > > On 10/8/07, Dirk Eddelbuettel <[EMAIL PROTECTED]> wrote: > > > > Were you doing this in a pbuilder chroot, or in a normal session? > > > > > > > > I have built numerous packages since I added the change, but all in > > > > pbuilder, > > > > and none may have had /usr/bin/java. > > > > > > I was doing it in a normal session (although under fakeroot); my > > > pbuilder chroot is set up for i386 and I haven't bothered changing it. > > > I wouldn't expect java to show up in a chroot normally anyway. > > > > That's the point of that change. In a pbuilder, I can safely run R CMD > > javareconf because the system that pbuilder run in is transient. > > > > On a 'live system', as you used, it is much harder. And I do not have a > > good > > answer for this. Simon invented R CMD javareconf because Java settings can > > change so damn much, adn R needs them 'normalised' to build Java packages. > > I am > > still not quite sure what to do here. > > > > Simon: Should I try to real surgery and try capture what 'R CMD javareconf > > -n' > > emits, and then export all that? I.e. get the benefit of the javareconf > > run > > without writing to /etc/R/Makeconf which I may not be allowed to touch? > > I'm wondering if there's some way to isolate the need to run R CMD > javareconf to packages that use Java when building... maybe a custom > version of R-cran.mk (R-cran-java.mk?) that is used when building > things that use Java?
Not a bad idea, but skits the issue. Even those would have to modify /etc/R/Makeconf when run the way we currently do. For the pbuilder case, I sort of did that as presence of /usr/bin/java would basically imply a Build-Depends on Java. > The only difficulty is trying to figure out what things need Java > automatically... parsing debian/control to find out if the package > needs java-something to build, possibly? R's DESCRIPTIOn would have a Depensd: rJava, I suppose. An explicit hook would be cleaner, but how do we get around the issue of reading/writing conf when we are non-root and don't want to alter a real Makeconf. A cheap out would be to have R read ./.Makeconf iff present, and we just write and then remove it. But it seems overkill to rejig R for this ... Dirk -- Three out of two people have difficulties with fractions. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
