# waiting on more info from me; see last paragraph tags 819998 + moreinfo thanks
Hi! Christian PERRIER wrote: > I suppose that this probably affects several packages and not this > one, only. If so, are there guidelines somewhere about Doing The Right > Thing? On kfreebsd we have a libusb2-dev, a different codebase that provides an interface generally compatible with libusb-1.0 (and others). Although it has a Provides: libusb-1.0-0-dev, that can only work if the Build-Depends: libusb-1.0-0-dev doesn't specify a version constrain, and in this case it does. This is quite common. So I typically recommend this: Build-Depends: libusb-1.0-0-dev (>= 1.0.9) [!kfreebsd-any], libusb2-dev [kfreebsd-any] and I perhaps should add this to some porting FAQ. Or, perhaps someone can suggest something better we could do than the Provides:. (Also note the comma between the alternative dependencies; a pipe does not do the obvious thing). For hurd, there is no USB stack yet, so... I guess there is no solution at the mooment. When there is, I'm not sure what the Build-Depends will need to be. With the above, it would stay BD-Uninstallable on hurd, waiting on libusb-1.0-0-dev, which is nice for porters to see clearly it is waiting on USB support. The out-of-date package can be decrufted on hurd-i386 if it is non-functional without USB. On kfreebsd, the next problem is that garmin-forerunner-tools ./configure uses /usr/bin/$(DEB_HOST_GNU_TYPE)-pkg-config and not the regular pkg-config. It scans only the multiarch path such as: /usr/lib/x86_64-linux-gnu/pkgconfig/libusb-1.0.pc but on kfreebsd it cannot find the libusb-1.0.pc provided by libusb2-dev: /usr/lib/pkgconfig/libusb-1.0.pc because we did not convert libusb2-dev to multiarched paths yet. This is the first time I'm seeing this; I suppose it is caused by --host=$(DEB_HOST_GNU_TYPE). I could multiarch libusb2-dev, but I worry some other packages do the opposite, and look in only /usr/lib/pkgconfig and not the multiarch paths... I think I should look into that, do some test rebuilds of reverse-deps in the archive, and then get back to you. Regards, -- Steven Chamberlain ste...@pyro.eu.org
signature.asc
Description: Digital signature