Rob Browning <[EMAIL PROTECTED]> writes: > - I tried building the package in merulo's unstable chroot. That > worked fine. Since the chroot doesn't have slib installed, and > neither did the buildd (according to the log), slib's not likely > to be the problem.
So the suggestion in the bug log was to actually *require* slib, not conflict with the old version. It's hard for me to see why that should make the build succeed, but who knows. > - The other suggestion was to make sure guile-1.6-libs isn't > installed. If the buildd starts from scratch for each package > (does it?), then that guile-1.6-libs couldn't be the problem. If > the buildd doesn't start from scratch, then it would seem a little > surprising if the problem only occurs on ia64. I agree that this is extremely unlikely. The only guess is that the bug log includes a purge of guile-1.6-libs; though even that shouldn't matter since guile-1.6-libs doesn't contain anything that would be left behind absent a purge. > If we could experiment in the buildd environment, it would be a lot > easier to figure out what's going on. Indeed. I have had major frustrations sometimes of this sort. One strategy is to use pbuilder instead of just an ordinary unstable chroot, since pbuilder is much closer to the sbuild reality than anything else. When I last had this kind of frustration, what finally helped was to put into debian/rules a complete printout of the configuration information after the run of configure, and then poring over that. Indeed, now that I've checked, the last time I had this kind of frustration, the problem was that lilypond behaved differently when stdin was a terminal and when it wasn't. This is perhaps one of the more weird differences between the sbuild/pbuilder environment and normal running of dpkg-buildpackage. Thomas -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
