The issue recently came up on IRC that many (many!) packages seem to have a Build-Depends: libc6-dev; so many, in fact, that it appears that the libc6.1-dev and libc0.3-dev packages both Provide: libc6-dev (which, I suppose, makes some sense, as they're all GNU libc, and thus, probably do have equivalent files). However, libc12-dev does *not* guarantee the exact same files, only "the C library files" (which, in effect, is probably the vast majority of what packages care about, in this case). It does provide libc12-dev, and it was suggested that it should just Provide libc6-dev as well, to avoid a mass bug filing. I assert that this is not a valid solution, as (unlike "variants of GNU libc" which almost certainly *do* have an equivalence), only libc-dev is proper, and providing libc6-dev would cause much confusion and breakage. So, two questions: 1) Should I file wishlist bugs against packages which fail to do something like "libc6-dev | libc-dev" when, in fact, they appear to build using only "generally found in libc" header files? Note that while it won't be all at once, there probably will be a number of these... it might be more appropriate to an automated check, but I suspect some things *do* actually depend on things from GNU libc. 2) 'libc6-dev' seems to have become the de facto 'GNU libc' check; Would it be useful for there be a pure virtual 'glibc-dev' shared among these packages? (This is, in the end, up to the glibc maintainer, really, but I figured it was worth asking while we were on the topic) -- *************************************************************************** Joel Baker System Administrator - lightbearer.com lucifer@lightbearer.com http://users.lightbearer.com/lucifer/
Attachment:
pgpDDNayVwDpT.pgp
Description: PGP signature