1 I would like to: represent the input tarballs as a
2 commit each (which all get merged together as if by git merge -s
3 subtree), and for quilt packages, each patch as a commit. But I want
4 to avoid (as much as possible) reimplementing the package extraction
5 algorithm in dpkg-source.
7 dpkg-source does not currently provide interfaces that look like they
8 are intended for what I want to do. And dgit wants to work with old
9 versions of dpkg, so I don't want to block on getting such interfaces
10 added (even supposing that a sane interface could be designed, which
13 So I intend to do as follows. (Please hold your nose.)
15 * dgit will untar each input tarball (other than the Debian tarball).
17 This will be done by scanning the .dsc for things whose names look
18 like (compressed) tarballs, and using the interfaces provided by
19 Dpkg::Compression to get at the tarball.
21 Each input tarball unpack will be done separately, and will be
22 followed by git-add and git-write tree, to obtain a git tree object
23 corresponding to the tarball contents.
25 That tree object will be made into a commit object with no parents.
26 (The package changelog will be searched for the earliest version
27 with the right upstream version component, and the information found
28 there used for the commit object's metadata.)
30 * For `3.0 (quilt), dgit will run
31 dpkg-source -x --skip-patches
33 git plumbing will be used to make the result into a tree and a
34 commit. The commit will have as parents all the tarballs previous
35 mentioned. The metadata will come from the .dsc and/or the final
38 dgit will then see if it has a series file. (dgit already rejects
39 packages with distro-specific series files, so we need worry only
40 about a single debian/patches/series file.)
42 If there is a series file, dgit will read it into memory. It will
43 then iterate over the series file, and each time:
44 - write into its playground a series file containing one
45 more non-comment non-empty line to previously
46 - run dpkg-source --before-build (which will apply that
48 - make git tree and commit objects, using the metadata from
49 the relevant patch file to make the commit (if available)
50 - each commit object has as a parent the previous commit
51 (either the previous commit, or the commit resulting from
54 After this the series file has been completely rewritten in
55 this way, the tree should be identical to the results of
58 * For source formats other than `3.0 (quilt)', dgit will do simply
61 Again, it will make that into a tree and a commit. (If this commit
62 has no changes in its tree from the first tarball commit. then the
63 two are squashed together.)
65 * As currently, there will be a final no-change-to-the-tree
66 pseudomerge commit which stitches the package into the relevant dgit
67 suite branch; ie something that looks as if it was made with git
70 * As currently, dgit will take steps so that none of the git trees
71 discussed above contain a .pc directory.
74 This has the following properties:
76 * Each input tarball is represented by a different commit; in usual
77 cases these commits will be the same for every upload of the same
80 * For `3.0 (quilt)' each patch's changes to the upstream files appears
81 as a single git commit (as is the effect of the debian tarball).
82 For `1.0' non-native, the effect of the diff is represented as a
83 commit. So eg `git blame' will show synthetic commits corresponding
84 to the correct parts of the input source package.
86 * It is possible to `git-cherry-pick' etc. commits representing `3.0
87 (quilt)' patches. It is even possible fish out the patch stack as
88 git branch and rebase it elsewhere etc., since the patch stack is
89 represented as a contiguous series of commits which make only the
90 relevant upstream changes.
92 * Every orig tarball in the source package is decompressed twice, but
93 disk space for only one extra copy of its unpacked contents is
94 needed. (The converse would be possible in principle but would be
95 very hard to arrange with the current interfaces provided by the
98 * No back doors into the innards of dpkg-source (nor changes to
99 dpkg-dev) are required.
101 * dgit does grow a dependency on Dpkg::Compression.
103 * Knowledge of the source format embedded in dgit is is restricted to
104 iterating over tarballs and manipulating debian/patches/series,
105 which dgit already does.
107 * dgit now depends on dpkg-source --before-build idempotently applying
108 patches as they successively appear on debian/patches/series.
110 * Perhaps the git commits generated by dgit to represent patches can
111 be made to round-trip nicely into tools like git-dpm and
114 I have found the information about tags in gbp-dch(1), but that
115 doesn't seem like it's applicable.
117 I have also found the information about tags in gbp-pq(1). From
118 that it looks like I ought to generate "Gbp-Pq: Name" and "Gbp-Pq:
121 * The scheme I describe avoids introducing a dependency from dgit to