This explains what's going on: <amck> joeyh: Been looking at #213834 (unnecessary libc-udeb). <joeyh> amck: any conclusions? <amck> The plan IIRC was to have explicit dependencies in all udebs; hence libc-udeb. <joeyh> well, we certianly want as good deps as we can, but with reduced libraries they kinda break down <amck> But we then pull in libs with mklibs ... I think thats the real breakage. <joeyh> the mklibs process acrtually generates fakeed entries in the status file <joeyh> libc-reduced or so <amck> mklibs should not touch libs, etc not explicity placed there by pkg-lists/dependencies. <amck> (So we can depend on dietlibc , etc instead in the future if we want). <joeyh> hmm, that's an interesting idea <joeyh> so you think that the build should install libc-udeb (from dependencies) and then mklibs should just bail if there is no libc or whatever there <joeyh> we'd need a slang-udeb then too, and some more <joeyh> and mklibs would need to be hacked <amck> We have a slang-udeb. <joeyh> no newt one though <joeyh> maybe discover too? <joeyh> no, there it is <amck> I'll look into it. Need to go quiet a baby. <amck> Good night all. -- see shy jo
Attachment:
signature.asc
Description: Digital signature