README.MultiArch

   1 Before we start with this topic: Note that MultiArch is not yet ready for
   2 prime time and/or for the casual user. The implementation is so far widely
   3 untested and only useful for developers of packagemanagment tools which
   4 use APT and his friends and maintainers of (upcoming) MultiArch packages.
   5 This README is especially NOT written for the casual user and is NOT a
   6 usage guide - you have been warned. It is assumed that the reader has
   7 at least a bit of knowledge about APT internals, dependency relations
   8 and the MultiArch spec [0].
   9
  10 Note also that the toolchain isn't ready yet, e.g. while you can simulate
  11 the installation of MultiArch packages they will more sooner than later
  12 cause enormous problems if really installed as dpkg can't handle MultiArch
  13 yet (no, --force-{overwrite,architecture} aren't good options here).
  14 Other parts of the big picture are missing and/or untested too.
  15 You have been warned!
  16
  17
  18 The implementation is focused on NOT breaking existing singleArch-only
  19 applications and/or systems as this is the current status-quo for all
  20 systems. Also, many systems don't need (or can't make use of) MultiArch,
  21 so APT will proceed in thinking SingleArch as long as it is not explicitly
  22 told to handle MultiArch:
  23 To activate MultiArch handling you need to specify architectures you
  24 want to be considered by APT with the config list APT::Architectures
  25 (Insert architectures in order of preference).
  26 APT will download Packages files for all these architectures in the
  27 update step. Exception: In the sourcelist is the optionfield used:
  28 deb [ arch=amd64,i386 ] http://example.org/ experimental main
  29 (This optionfield is a NOP in previous apt versions)
  30
  31 Internally in APT a package is represented as a PkgIterator -
  32 before MultiArch this PkgIterator was architecture unaware,
  33 only VerIterators include the architecture they came from.
  34 This is/was a big problem as all versions in a package are
  35 considered for dependency resolution, so pinning will not work in all cases.
  36
  37 The problem is solved by a conceptional change:
  38 A PkgIterator is now architecture aware, so the packages
  39 of foobar for amd64 and for i386 are now for apt internal totally
  40 different packages. That is a good thing for e.g. pinning, but
  41 sometimes you need the information that such packages are belonging together:
  42 All these foobar packages therefore form a Group accessible with GrpIterators.
  43 Note that the GrpIterator has the same name as all the packages in this group,
  44 so e.g. apt-cache pkgnames iterates over GrpIterator to get the package names:
  45 This is compatible to SingleArch as a Group consists only of a single package
  46 and also to MultiArch as a Group consists of possible many packages which
  47 all have the same name and are therefore out of interest for pkgnames.
  48
  49
  50 Caused by the paragraph "Dependencies involving Architecture: all packages"
  51 in the MultiArch spec we have a second major conceptional change
  52 which could even break existing applications, but we hope for the best…
  53 An Architecture: all package is internally split into pseudo packages
  54 for all MultiArch Architectures and additional a package with the
  55 architecture "all" with no dependencies which is a dependency of all
  56 these architecture depending packages. While the architecture depending
  57 packages are mainly used for dependency resolution (a package of arch A which
  58 depends on an arch all package assumes that the dependencies of this package
  59 are also from arch A. Packages also sometimes change from any to all or v.v.)
  60 the arch "all" package is used for scheduling download/installation of the
  61 underlying "real" package. Note that the architecture depending packages can
  62 be detected with Pseudo() while the "all" package reports exactly this arch
  63 as package architecture and as pseudo architecture of the versions of this pkg.
  64 Beware: All versions of a "real" architecture all package will be report "all"
  65 as their architecture if asked with Arch() regardless if they are the "all" or
  66 the architecture depending packages. If you want to know the architecture this
  67 pseudo package was created for call Arch(true). Also, while the spec say that
  68 arch:all packages are not allowed to have a MultiArch flag APT assigns a
  69 special value to them: MultiArch: all.
  70
  71
  72 As you might guess this arch:all handling has a few problems (but we think so
  73 far that the problems are minor compared to the problems we would have with
  74 other implementations.)
  75 APT doesn't know which pseudo packages of such an arch all package are
  76 "installed" (to satisfy dependencies), so APT will generate a Cache in which
  77 all these pseudo packages are installed (e.g. apt-cache policy will display
  78 them all as installed). Later in the DepCache step it will "remove"
  79 all pseudo packages whose dependencies are not satisfied.
  80 The expense is that if the package state is broken APT could come to the
  81 conclusion to "remove" too many pseudo packages, but in a stable environment
  82 APT should never end up in a broken system state…
  83
  84
  85 Given all these internal changes it is quite interesting that the actual
  86 implementation of MultiArch is trivial: Some implicit dependencies and a few
  87 more provides are all changes needed to get it working. Especially noteworthy
  88 is that it wasn't needed to change the resolver in any way and other parts only
  89 need to be told about ignoring pseudo packages or using GrpIterator instead of
  90 PkgIterator, so chances are good that libapt-applications will proceed to work
  91 without or at least only require minor changes, but your mileage may vary…
  92
  93
  94 Known Issues and/or noteworthy stuff:
  95 * The implementation is mostly untested, so it is very likely that APT will
  96   eat your kids if you aren't as lucky as the author of these patches.
  97 * the (install)size of a pseudo package is always NULL - if you want to know
  98   the (install)size you need to get the info from the arch "all" package.
  99 * It is maybe confusing, but the arch "all" package does have the same versions
 100   and in general roughly the same information with one subtil difference:
 101   It doesn't have any dependency, regardless of the type. The pseudo packages
 102   depend on this package.
 103 * apt-cache policy foobar on installed architecture all package foobar will
 104   report all architecture depending packages as installed. Displaying here the
 105   correct information would require to build the complete DepCache…
 106 * [BUG] An installed package which changes the architecture from any to all
 107   (and v.v.) shows up in the NEW packages section instead of UPGRADE.
 108 * [TODO] Investigate the DepCache pseudo-package killer heuristic:
 109   e.g. add more safety guards…
 110 * [FIXME] a few corner cases/missing features marked as FIXME in the code
 111
 112
 113 [0] https://wiki.ubuntu.com/MultiarchSpec