On Sun, May 17, 2009 at 05:30:39AM +0200, Guillem Jover wrote: > On Sun, 2009-05-17 at 04:26:02 +0200, Guillem Jover wrote: > > On Sat, 2009-05-16 at 22:24:43 +0200, Bill Allombert wrote: > > > On Sat, May 16, 2009 at 09:54:35PM +0200, Norbert Preining wrote: > > > > On Sat, 16 May 2009, Bill Allombert wrote: > > > > > Does Lenny includes this trigger ? > > > > > > > > > Did you consider what will when a partial lenny to squeeze update is > > > > > made ? > > > > Yes, most of the plan has been drafted with the idea of inflicting > > minimal disruption. > > > > > > Yes. If the new texinfo/install-info is installed it will work on the > > > > triggers. Packages that are old will call install-info which is a > > > > wrapper > > > > and will do nothing. If the old install-info/texinfo is installed then > > > > the normal install-info procedure is called. > > > > > > What happens if a new package is installed but not the new install-info ? > > > > The info files are not usually ‘readable’ w/o an info-reader. The new > > info-reader Providing packages will Depend on the install-info package, > > so they will get a generated dir when installed. And the first dpkg > > version to stop shipping the real install-info will Break all old > > info-readers versions not Depending on the install-info package. Does > > this resolve your concerns? > > Actually, no, I guess it does not. In case the user has no upgraded > dpkg nor any of the info-readers, the user could upgrade a info > providing package and that would not call install-info anymore (in > this particular case things would probably just work, as the info dir > section is already on the dir file, but not for new info files, or > renamed files).
Yes this was my point. > So there's several options that come to mind for that: > > * We don't care, and expect users might miss docs on the dir file in > some cases or need to upgrade dpkg or any of the info-readers. > * Make info providing packages depend on install-info. > * Make info providing packages Break old dpkgs. > * Not remove calls to install-info from packages until squeeze+1 > (and make install-info wrappers not warn in some conditions). > > Probably the sanest and safest is the last one, but slowest and with > less immediate benefits. OTOH not registering some docs on the dir > file is not that grave, as they will get readded whe upgrading. > So I'd go for the "we don't care", but would not mind being more > conservative. Since all packages that use install-info need to be changed, options 2) seems doable, and since install-info used to be provided by dpkg it even makes sense. I do not have experience with the behaviour of Break during upgrade (with aprt or aptitude) to comment on 3) The point is that we are introducing a potential breakage during upgrade. In this instance it is rather trivial, but when triggers are used for more critical things like boot-loaders, this should be documented and prevented. Cheers, -- Bill. <ballo...@debian.org> Imagine a large red swirl here. -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org