Jordi Pujol wrote: > now the default behaviour for update-rc.d with option "defaults" is parse > the > LSB Headers, therefore we can leave blank the options for dh_installlinit and > it will pass the option "defaults" to update-rc.d, > but when update-rc.d works in legacy mode, dh_installinit should write the > start and stop values, obtained from parsing the LSB headers.
If the LSB Default-Start and Default-Stop headers can be trusted, why does "update-rc.d defaults" not already parse and use those in legacy mode? I don't like adding yet another parser for this, and it seems it should be up to update-rc.d to decide whether to use those headers or not. My suspicion is that, since they are unused, they probably are less correct than the default "defaults" behavior in most cases, and that by using this unverified infomation source, debhelper would break a lot of packages. Can you prove otherwise? It does seem that by making most systems use dependency based boot, while leaving a fraction in legacy mode, we now have a situation where the legacy mode is going to tend to bit-rot, since it is a) hard to maintain and b) won't get wide use and testing. But I don't see this change to debhelper signifiantly ammeloriating that problem, and it could easily make it much worse. -- see shy jo
signature.asc
Description: Digital signature
