Bug#198569: [ITP]: r-noncran-design -- Regression modeling strategies
On Jun 24, Dirk Eddelbuettel wrote:
> Not sure I follow. I did the following:
> [snipped because post-mode decided to kill your entire train of
> thought here... grr]
I think there's a nasty issue that will crop up if you try to work
from the files in the Debian archive. debuild -us -uc, then wipe your
build directories so all you have left are design_1.3.1.orig.tar.gz,
design_1.3.1-1.dsc and design_1.3.1-1.diff.gz
Now try dpkg-source -x design_1.3.1-1.dsc. This will untar everything
into a directory "design-1.3.1-1," not "Design". cd in there and
debuild -us -uc again.
If that build process produces /usr/lib/R/site-library/Design rather
than /usr/lib/R/site-library/design-1.3.1-1, I'll be very surprised.
What this will produce is packages that will build OK on your box, but
the buildds will do exactly what I outlined and produce ports with
broken .debs. It won't fail to build from source, but if you check
the buildd log, it will be wrong and will produce exactly the problems
Doug describes. You just won't see them since i386 will be fine,
since you built it yourself and knew that the directory wasn't
supposed to be called design-1.3.1-1; the buildd doesn't know any
better, malheureusement :-/
For example, see (sorry for linewrap):
http://buildd.debian.org/fetch.php?&pkg=tseries&ver=0.9.12-2&arch=m68k&stamp=1056336148&file=log&as=raw
Skip to the end and scroll up to where it runs dpkg -c on the built
deb. (Or you can just wget the file from the pool and dpkg -c it.)
Hence the ad-hackery that I did in mcmcpack and coda (though, since
coda is an arch: all package, it doesn't go to the buildds normally).
Chris
--
Chris Lawrence <lawrencc@debian.org> - http://blog.lordsutch.com/
Reply to: