Debian package management

dpkg is the low level

dpkg installs, removes and queries individual package files. It knows about dependencies in the sense that it refuses to install a package whose dependencies are missing — but it will not go and fetch them, because it has no idea where packages come from. dpkg -i installs a file, dpkg -r removes a package but keeps its configuration, and dpkg -P purges it including configuration. dpkg -l lists installed packages, dpkg -L lists what a package installed, and dpkg -S finds which package owns a given file.

dpkg-reconfigure re-runs a package's configuration questions after installation, which is the answer whenever a question asks how to change something you were asked during install. dpkg --get-selections and --set-selections copy the set of installed packages from one machine to another.

A second group of operations works on a .deb FILE rather than on the installed database, and the exam keeps them apart from the query flags for exactly that reason. A .deb is an archive with a control member and a data member, and dpkg -c (long form --contents) lists the data member — every file the package WOULD place on the system — without installing any of it, while dpkg -e and dpkg -x extract the control files or the data. All of them are really dpkg-deb, which dpkg hands the request straight to, so dpkg-deb -c htop.deb and dpkg -c htop.deb are one command under two names, and dpkg-deb -I prints the control information. The pair that reads alike and is not alike: dpkg -L lists the files of an INSTALLED package, dpkg -c lists the contents of a file sitting on disk.

APT is the level that knows where packages live

APT reads /etc/apt/sources.list and the files in /etc/apt/sources.list.d/, downloads package indexes with apt-get update, and resolves dependencies before installing. apt-get install fetches and installs with everything it needs; apt-get upgrade upgrades installed packages without removing any; apt-get dist-upgrade (or apt full-upgrade) is allowed to add and remove packages to satisfy a larger change. apt-cache search and apt-cache show query the indexes, and apt-cache depends walks a dependency tree.

The apt command is the newer, friendlier front end over the same machinery — apt install, apt search, apt list --installed — intended for interactive use. aptitude is a third front end with its own dependency resolver and a full-screen interface. All three ultimately act on the same dpkg database.

There are two file-to-package searches on this family and they differ in scope, which is the whole question. dpkg -S reads the file lists under /var/lib/dpkg/info, so it can only answer about packages already installed and says nothing at all about a header like /usr/include/zlib.h that is missing precisely because its package is not there. apt-file search /usr/include/zlib.h searches the Contents indexes the repositories publish, so it names the package to install; it is a separate package on most systems, and its index has to be fetched with apt-file update before the first search. dnf provides is the same question on the RPM side.