On Tue, Aug 07, 2007 at 11:55:54AM +0200, Loïc Minier <[EMAIL PROTECTED]> wrote: > Hi, > > On Tue, Aug 07, 2007, Neil Williams wrote: > > the -dbg (with SONAME) > [...] > > provide a binary package > > librarynamesoversion-dbg > [...] > > Josh Triplett <[EMAIL PROTECTED]> on 2007-04-22 19:48 > > Please don't impose "librarynamesoversion-dbg", a lot of source > packages do not only provide a library, but also programs, and we > shouldn't duplicate all packages into having a -dbg (name-dbg). > This would also imply having multiple -dbg packages when there are two > libraries in the same source. > > One example of such a package could be nautilus which has a > nautilus-dbg with debugging symbols for nautilus and > libnautilus-extension. > > I recommend naming the -dbg $source-dbg instead.
That wouldn't work with libraries that have two different flavours that conflict with each other. I actually think -dbg packages should all be removed, and replaced with special packages, such as .ddeb. There would be a csimple .deb <-> .ddeb mapping. I also want to investigate how feasible (and useful) it would be to have "mini" core dumps that could be sent by users when reporting crash bugs without the need to install debugging package, and having tools for maintainers to actually get a significan backtrace from these informations and .ddeb packages. The first part of this whole reflection I was having was to check how many packages don't build with debugging symbols. I was considering running a poll quite soonish, to get input from package maintainers about the informations they need most of the time or would need from bug reports (basis of a "do we really need *all* debugging information" reflection). Obviously, there is still a long way... http://bugs.debian.org/cgi-bin/[EMAIL PROTECTED];which=tag&data=nostrip Mike -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
