devuan's apt repo merger
git clone git://
Log | Files | Refs | | LICENSE

commit 7fdbb28a7fb4b11adfbdc25356bf006b28edebc4
parent 5d7bdbb57bcf9c78ae13e79c7550d2a120dfc1ce
Author: parazyd <>
Date:   Wed,  7 Jun 2017 19:38:22 +0200

add TODO

Diffstat: | 8++++++++
doc/ | 50--------------------------------------------------
2 files changed, 8 insertions(+), 50 deletions(-)

diff --git a/ b/ @@ -0,0 +1,8 @@ +amprolla TODO +============= + +* Implement proper logging +* Log when there is a newer version in a repository of a lesser priority +* Translation files +* Contents files +* Investigate why detached GPG signatures end up broken diff --git a/doc/ b/doc/ @@ -1,50 +0,0 @@ -* an older spool directory hierarchy, for testing the merging: - (2GB) - - -workflow --------- - -`amprolla-init` creates the `spool` directory hierarchy. This hierarchy -is the untouched repositories we want to merge. They can be found in -`lib/` in the `repos` dict. - -`fs.crawl()` takes the info from `` and generates the paths we -use to figure out what to download. Then `popDirs()` in `amprolla-init` -uses these paths to actually create the directory structure and download -the `Release` files from there. (NOTE: at a final point it should also -check the gpg signatures from the `InRelease` file) - -After the Release files are downloaded, they are parsed for their -contents and all the files they list are downloaded in series (possibly -should be parallel). - -So now we have all the files we need to create a merge. - - - -The merge will be done in the `merged/` directory. To do a successful -merge, these are the points that need to be accomplished: - - * we must skip packages that are in `` - * for each `binary-$arch` we must create a new `Packages` file, - containing our merge: - * we parse every package and fill in a dict - * first priority 0, then 1, then 2, etc... - * if a package already exists from a higher-priority repo - it - gets skipped - * (NOT SURE): if a package from `banpkgs` is in a package's - dependency list - that package gets skipped - * once we've finished the iteration, we dump a new Packages file - from the updated dict - - - -After the initial merge, we need to watch for updates. My idea is to -make amprolla pool the above-mentioned Release files as they contain -enough metadata for us to find out what changed. They also contain a -date entry so we can see if there was actually an update without digging -too deep. - -So if we figure out there was an update, we download the new file, parse -it into a dict and compare it to the old version of that file/dict.