1 Towards replacing rpm + yum (0.1):
5 - pre install check; check that dirs can be created (no files where
6 want to create dirs), move config files according to file
9 - run hooks; probably fork to shell and in the parent just do a
10 write() of the hook section to a pipe to the shell.
12 - store rpm headers for installed packages.
14 - versions in depsolving
18 - merge file lists when merging package sets
22 - figure out how to canonically represent empty string... ~0?
24 - space calculation before transaction, but ideally, do a number of
27 - pre-link changing binaries and libs on disk screwing up checksum?
29 - nail down byte-order and word sizes of repo file.
31 - version the sections in the file, put the element size in the header
32 so we can add stuff to elements in a backwards compatible way.
33 maybe not necessary, we can just add sections that augment the
34 sections we want to add to (similar to how rpm has add versioned
37 - pipelined download and install; topo-sort packages in update set,
38 pick one with all deps in the current set, add it to the current set
39 and satisfy deps against update set => result: minimal update
40 transaction. Queue download and install/update transaction for the
41 packages in the minimal set, start over. This also makes the
42 installation phase much more interruptible, basically just stop
43 after a sub-transaction finishes. As we keep the update set around
44 as a target, we can restart if needed. Probably don't need to, can
45 just do a new update. During a sub-transaction we should keep the
46 target set (i.e. the current set to be) around as a lock file
47 (system.repo.lock or so, see git) so that razor updates are
48 prevented if the systems crashes during an update.
52 - keep history of installed packages/journal of package transaction,
53 so we can roll back to yesterday, or see what got installed in the
56 - transactions, proper recovery, make sure we don't poop our package
57 database (no more rm /var/lib/rpm/__cache*).
59 - rewrite qsort and bsearch that doesn't require global context var
60 and can output a map describing the permutaion.
62 - use hash table for package and property lists so we only store
63 unique lists (like for string pool).
65 - use existing, running system as repo; eg
67 razor update razor://other-box.local evince
69 to pull eg the latest evince and dependencies from another box. We
70 should be able to regenerate a rzr pkg from the system so we can
71 reuse the signature from the originating repo.
73 - Ok, maybe the fastest package set merge method in the end is to use
74 the razor_importer, but use a hash table for the properties. This
75 way we can assign them unique IDs immediately (like tokenizing
78 - test suite should be easy, just keep .repo files around and test
79 different type of upgrades that way (obsoletes, conflicts, file
80 conflicts, file/dir problems etc). Or maybe just keep a simple file
81 format ad use a custom importer to create the .repo files.
85 do { ... } while (((e++)->name & RAZOR_ENTRY_LAST) == 0);
87 idiom for iteration of directories.
89 - overlay package sets? mount a read-only /usr over nfs or from the
90 virt-host and have a local package set overlaid over the read-only
91 one. shouldn't need new features in the core package set data
92 structure, but should be just conventions on top. we have the base
93 package set from the r/o system, the overlay set from the local
94 system and we can have an effective package set which is the merge
95 of everything from the overlay into the base set. the effective set
96 is easy to compute and we could do it on the fly or cache it. or
97 maybe the effective set is the on-disk representation and the
98 overlay can be computed when needed, we just keep a link back to the
101 - incremental rawhide repo updates? instead of downloading 10MB zipped
102 repo every time, download a diff repo? Should be pretty small,
103 especially if we don't have file checksums in metadata. Filenames
104 and properties are for the most part already present, typically just
105 a version bump plus maybe tweaking a couple requires. The upstream
106 repo can store multiple incremental updates in one big file and
107 provide an index file that maps updates for a given date (we should
108 use repo-file checksums though) to a range in the file: Download the
109 index file, search for a match for your latest rawhide.repo file,
110 download range of updates that brings it up to date.
112 - use hash tables for dirs when importing files to avoid qsorting all
117 - eliminate duplicate entries in package property lists.
119 - corner cases such as no files/properties in repo etc segfault.