On Sat, 27 Nov 2010 13:26:46 +0000 Luke Kenneth Casson Leighton <lkcl@lkcl.net> wrote: > -T<target> call debian/rules <target> with the proper > environment > > > rrright - am i correct in thinking that if you do "dpkg-buildpackage > -rfakeroot -Tfoo" it does _nothing_ but run that particular task in > the debian/rules makefile? assuming that the task's dependencies (in > the debian/rules aka makefile) have all been satisfied up to that > point, that is. It will ensure that the dependent targets have been met. Depending on what you are doing, you may need to stop dpkg-bp doing the initial clean, yet not calling clean is also not a good idea. > if so, it would almost be a trivial job to back-end the bitbake > "tasks" concept into the dpkg-buildpackage "target" conecept, Running individual rules is ONLY suitable for debugging the actual packaging. You cannot build Debian packages using individual targets in isolation - the results will not be reproducible, will not work as the normal package should and will result in obscure bugs. These packages need to be uploaded as binaries for other people to install on their own systems - that's the point, you're trying to mangle a source-based method to a binary-based method. Give it up. Debian packages for distribution MUST be built in a clean chroot without any mangling or hacking - calling the debian/rules as a complete set, in a single process. (and yes, you MUST produce individual binaries, not just a 'complete' image.) > one-for-one compile, configure, install, package, sign. in between > you'd perform special cross-compile "tweaks" which are well-known > already in the openembedded environment and are prooobably similar to > what dpkg-cross does.... *scratches head*... Not even slightly. > is this easy to undeststand? Yes, it's easy to understand that it's the wrong approach. > i'm a little confused at the lack of > response and discussion: if i haven't made this clear enough i > apologise - please do tell me what i need to say so that this can be > discussed. The two approaches are incompatible, that's why nobody is taking this on. The idea of a hybrid is a dead-end - please stop toiling/trolling about it. -- Neil Williams ============= http://www.linux.codehelp.co.uk/
Attachment:
pgpMeKFxfQYMC.pgp
Description: PGP signature