Compare commits

...

15 commits

Author SHA1 Message Date
Bruno BELANYI efbfc3f4a0 WIP: migrate to full featured filter list
I don't like the fact that ordering is significant, given that we would
be processing this filter list in multiple places...
2021-04-18 13:36:47 +00:00
Antoine Martin d942dc1098 ci: setup cachix build 2021-04-08 04:24:48 +02:00
Antoine Martin bc32f8a565 README: add note about missing versions on crates.io 2021-04-08 02:49:19 +02:00
Antoine Martin 6c4a496ede lohr: v0.4.0
This is the first lohr release that compiles on stable Rust, yay!
2021-04-08 02:49:19 +02:00
Antoine Martin 264b7baccd cargo: bump dependencies 2021-04-08 01:51:07 +02:00
Antoine Martin 776876233f gitea: use ssh_url instead of clone_url
Fixes #2
2021-04-08 01:46:52 +02:00
Bruno BELANYI 3fedfa3342 nix: flake: build package with vanilla 'nixpkgs' 2021-04-08 01:41:15 +02:00
Antoine Martin 378823a1fb ci: switch to stable toolchain 2021-04-08 01:41:15 +02:00
Antoine Martin 56f82c6467 README: remove mention of nightly compiler 2021-04-08 01:41:12 +02:00
Antoine Martin 3a93d9f994 lohr: build on rust stable
Switching to Rocket's master branch (soon to be 0.5.0) allows building
with a stable Rust compiler, yay!
2021-04-08 01:39:21 +02:00
Antoine Martin da20e2c9ac cargo fmt
whoops
2021-03-31 21:04:30 +02:00
Antoine Martin 29c0a4abdf lohr: v0.3.3 2021-03-31 19:39:04 +02:00
Antoine Martin 1506ce1dac README: switch to Markdown
Unfortunately crates.io doesn't show Org READMEs, and GitHub's Org-mode
renderer isn't amazing either. So might as well switch to a well
supported format, even though I liked Org better.
2021-03-31 19:28:24 +02:00
Antoine Martin fe52787f4d lohr: v0.3.2 2021-03-31 19:06:36 +02:00
Antoine Martin e68fcf3e52 settings: blacklist isn't required 2021-03-31 18:57:50 +02:00
14 changed files with 1213 additions and 466 deletions

19
.github/workflows/cachix.yaml vendored Normal file
View file

@ -0,0 +1,19 @@
name: "Cachix build"
on: [push]
jobs:
build:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v2.3.4
- uses: cachix/install-nix-action@v13
with:
install_url: https://nixos-nix-install-tests.cachix.org/serve/lb41az54kzk6j12p81br4bczary7m145/install
install_options: '--tarball-url-prefix https://nixos-nix-install-tests.cachix.org/serve'
extra_nix_config: |
experimental-features = nix-command flakes
- uses: cachix/cachix-action@v8
with:
name: alarsyo
authToken: '${{ secrets.CACHIX_AUTH_TOKEN }}'
- run: |
nix build --verbose

View file

@ -10,11 +10,11 @@ jobs:
- name: Checkout sources
uses: actions/checkout@v2
- name: Install nightly toolchain
- name: Install stable toolchain
uses: actions-rs/toolchain@v1
with:
profile: minimal
toolchain: nightly
toolchain: stable
override: true
- name: Run cargo check
@ -29,11 +29,11 @@ jobs:
- name: Checkout sources
uses: actions/checkout@v2
- name: Install nightly toolchain
- name: Install stable toolchain
uses: actions-rs/toolchain@v1
with:
profile: minimal
toolchain: nightly
toolchain: stable
override: true
- name: Run cargo test
@ -48,11 +48,11 @@ jobs:
- name: Checkout sources
uses: actions/checkout@v2
- name: Install nightly toolchain
- name: Install stable toolchain
uses: actions-rs/toolchain@v1
with:
profile: minimal
toolchain: nightly
toolchain: stable
override: true
components: rustfmt, clippy

1153
Cargo.lock generated

File diff suppressed because it is too large Load diff

View file

@ -1,6 +1,6 @@
[package]
name = "lohr"
version = "0.3.1"
version = "0.4.0"
authors = ["Antoine Martin <antoine@alarsyo.net>"]
edition = "2018"
license = "Apache-2.0 OR MIT"
@ -23,7 +23,7 @@ serde_yaml = "0.8.17"
sha2 = "0.9.3"
[dependencies.rocket]
version = "0.4.7"
version = "0.5.0-dev"
# don't need private-cookies
default-features = false
@ -31,3 +31,6 @@ default-features = false
version = "2.33.3"
# no need for suggestions or color with only one argument
default-features = false
[patch.crates-io]
rocket = { git = "https://github.com/SergioBenitez/Rocket", rev = "8d4d01106e2e10b08100805d40bfa19a7357e900" }

138
README.md Normal file
View file

@ -0,0 +1,138 @@
# lohr
`lohr` is a Git mirroring tool.
I created it to solve a simple problem I had: I host my own git server at
<https://git.alarsyo.net>, but want to mirror my public projects to GitHub /
GitLab, for backup and visibility purposes.
GitLab has a mirroring setting, but it doesn't allow for multiple mirrors, as
far as I know. I also wanted my instance to be the single source of truth.
## How it works
Gitea is setup to send webhooks to my `lohr` server on every push update. When
`lohr` receives a push, it clones the concerned repository, or updates it if
already cloned. Then it pushes the update to **all remotes listed** in the
[.lohr](.lohr) file at the repo root.
### Destructive
This is a very destructive process: anything removed from the single source of
truth is effectively removed from any mirror as well.
## Installing
`lohr` is [published on crates.io](https://crates.io/crates/lohr), so you can
install it with `cargo install`:
$ cargo install lohr
Note: currently this method won't get you the latest version of `lohr`, as it
depends on Rocket v0.5.0, which isn't released yet. Updated versions of `lohr`
will be published on crates.io as soon as Rocket v0.5.0 releases.
## Setup
### Quickstart
Setting up `lohr` should be quite simple:
1. Create a `Rocket.toml` file and [add your
configuration](https://rocket.rs/v0.4/guide/configuration/).
2. Export a secret variable:
$ export LOHR_SECRET=42 # please don't use this secret
3. Run `lohr`:
$ cargo run # or `cargo run --release` for production usage
4. Configure your favorite git server to send a webhook to `lohr`'s address on
every push event.
I used [Gitea's webhooks format](https://docs.gitea.io/en-us/webhooks/), but
I **think** they're similar to GitHub and GitLab's webhooks, so these should
work too! (If they don't, **please** file an issue!)
Don't forget to set the webhook secret to the one you chose above.
5. Add a `.lohr` file containing the remotes you want to mirror this repo to:
git@github.com:you/your_repo
and push it. That's it! `lohr` is mirroring your repo now.
### Configuration
#### Home directory
`lohr` needs a place to clone repos and store its data. By default, it's the
current directory, but you can set the `LOHR_HOME` environment variable to
customize it.
#### Shared secret
As shown in the quickstart guide, you **must** set the `LOHR_SECRET` environment
variable.
#### Extra remote configuration
You can provide `lohr` with a YAML file containing additional configuration. You
can pass its path to the `--config` flag when launching `lohr`. If no
configuration is provided via a CLI flag, `lohr` will check the `LOHR_CONFIG`
environment variable. If the environment variable isn't set either, it will
check in `LOHR_HOME` is a `lohr-config.yaml` file exists, and try to load it.
This file takes the following format:
``` yaml
default_remotes:
- "git@github:user"
- "git@gitlab:user"
additional_remotes:
- "git@git.sr.ht:~user"
filters:
- FIXME:
```
- `default_remotes` is a list of remotes to use if no `.lohr` file is found in a
repository.
- `additional_remotes` is a list of remotes to add in any case, whether the
original set of remotes is set via `default_remotes` or via a `.lohr` file.
- `blacklist` is a list of regular expressions to match against the full
repository names. Any that matches will not be mirrored, even if it contains a
`.lohr` file.
Both settings take as input a list of "stems", i.e. incomplete remote addresses,
to which the repo's name will be appended (so for example, if my
`default_remotes` contains `git@github.com:alarsyo`, and a push event webhook is
received for repository `git@gitlab.com:some/long/path/repo_name`, then the
mirror destination will be `git@github.com:alarsyo/repo_name`.
## Contributing
I accept patches anywhere! Feel free to [open a GitHub Pull
Request](https://github.com/alarsyo/lohr/pulls), [a GitLab Merge
Request](https://gitlab.com/alarsyo/lohr/-/merge_requests), or [send me a patch
by email](https://lists.sr.ht/~alarsyo/lohr-dev)!
## Why lohr?
I was looking for a cool name, and thought about the Magic Mirror in Snow White.
Some **[furious wikipedia
searching](https://en.wikipedia.org/wiki/Magic_Mirror_(Snow_White))** later, I
found that the Magic Mirror was probably inspired by [the Talking Mirror in Lohr
am Main](http://spessartmuseum.de/seiten/schneewittchen_engl.html). That's it,
that's the story.
## License
`lohr` is distributed under the terms of both the MIT license and the Apache
License (Version 2.0).
See [LICENSE-APACHE](LICENSE-APACHE) and [LICENSE-MIT](LICENSE-MIT) for details.

View file

@ -1,143 +0,0 @@
#+title: lohr
=lohr= is a Git mirroring tool.
I created it to solve a simple problem I had: I host my own git server at
[[https://git.alarsyo.net]], but want to mirror my public projects to GitHub /
GitLab, for backup and visibility purposes.
GitLab has a mirroring setting, but it doesn't allow for multiple mirrors, as
far as I know. I also wanted my instance to be the single source of truth.
** How it works
Gitea is setup to send webhooks to my =lohr= server on every push update. When
=lohr= receives a push, it clones the concerned repository, or updates it if
already cloned. Then it pushes the update to *all remotes listed* in the [[file:.lohr][.lohr]]
file at the repo root.
*** Destructive
This is a very destructive process: anything removed from the single source of
truth is effectively removed from any mirror as well.
** Installing
=lohr= is [[https://crates.io/crates/lohr][published on crates.io]], so you can install it with ~cargo install~:
#+begin_src sh
$ cargo +nightly install lohr
#+end_src
We currently require a nightly compiler because [[https://github.com/SergioBenitez/Rocket][Rocket]] needs one to compile (a
0.5.0 which compiles on stable Rust is in the making, stay tuned!). You can
install a nightly toolchain with the following command:
#+begin_src sh
$ rustup install nightly
#+end_src
** Setup
*** Quickstart
Setting up =lohr= should be quite simple:
1. Create a =Rocket.toml= file and [[https://rocket.rs/v0.4/guide/configuration/][add your configuration]].
2. Export a secret variable:
#+begin_src sh
$ export LOHR_SECRET=42 # please don't use this secret
#+end_src
3. Run =lohr=:
#+begin_src sh
$ cargo run # or `cargo run --release` for production usage
#+end_src
4. Configure your favorite git server to send a webhook to =lohr='s address on
every push event.
I used [[https://docs.gitea.io/en-us/webhooks/][Gitea's webhooks format]], but I *think* they're similar to GitHub and
GitLab's webhooks, so these should work too! (If they don't, *please* file an
issue!)
Don't forget to set the webhook secret to the one you chose above.
5. Add a =.lohr= file containing the remotes you want to mirror this repo to:
#+begin_example
git@github.com:you/your_repo
#+end_example
and push it. That's it! =lohr= is mirroring your repo now.
*** Configuration
**** Home directory
=lohr= needs a place to clone repos and store its data. By default, it's the
current directory, but you can set the =LOHR_HOME= environment variable to
customize it.
**** Shared secret
As shown in the quickstart guide, you *must* set the =LOHR_SECRET= environment
variable.
**** Extra remote configuration
You can provide =lohr= with a YAML file containing additional configuration. You
can pass its path to the =--config= flag when launching =lohr=. If no
configuration is provided via a CLI flag, =lohr= will check the =LOHR_CONFIG=
environment variable. If the environment variable isn't set either, it will
check in =LOHR_HOME= is a =lohr-config.yaml= file exists, and try to load it.
This file takes the following format:
#+begin_src yaml
default_remotes:
- "git@github:user"
- "git@gitlab:user"
additional_remotes:
- "git@git.sr.ht:~user"
blacklist:
- "private-.*"
#+end_src
- ~default_remotes~ is a list of remotes to use if no ~.lohr~ file is found in a
repository.
- ~additional_remotes~ is a list of remotes to add in any case, whether the
original set of remotes is set via ~default_remotes~ or via a =.lohr= file.
- ~blacklist~ is a list of regular expressions to match against the full
repository names. Any that matches will not be mirrored, even if it contains a
`.lohr` file.
Both settings take as input a list of "stems", i.e. incomplete remote addresses,
to which the repo's name will be appended (so for example, if my
~default_remotes~ contains ~git@github.com:alarsyo~, and a push event webhook
is received for repository =git@gitlab.com:some/long/path/repo_name=, then the
mirror destination will be =git@github.com:alarsyo/repo_name=.
** Contributing
I accept patches anywhere! Feel free to [[https://github.com/alarsyo/lohr/pulls][open a GitHub Pull Request]], [[https://gitlab.com/alarsyo/lohr/-/merge_requests][a GitLab
Merge Request]], or [[https://lists.sr.ht/~alarsyo/lohr-dev][send me a patch by email]]!
** Why lohr?
I was looking for a cool name, and thought about the Magic Mirror in Snow White.
Some *[[https://en.wikipedia.org/wiki/Magic_Mirror_(Snow_White)][furious wikipedia searching]]* later, I found that the Magic Mirror was
probably inspired by [[http://spessartmuseum.de/seiten/schneewittchen_engl.html][the Talking Mirror in Lohr am Main]]. That's it, that's the
story.
** License
=lohr= is distributed under the terms of both the MIT license and the Apache
License (Version 2.0).
See [[file:LICENSE-APACHE][LICENSE-APACHE]] and [[file:LICENSE-MIT][LICENSE-MIT]] for details.

View file

@ -31,43 +31,6 @@
"type": "github"
}
},
"mozillapkgs": {
"flake": false,
"locked": {
"lastModified": 1603906276,
"narHash": "sha256-RsNPnEKd7BcogwkqhaV5kI/HuNC4flH/OQCC/4W5y/8=",
"owner": "mozilla",
"repo": "nixpkgs-mozilla",
"rev": "8c007b60731c07dd7a052cce508de3bb1ae849b4",
"type": "github"
},
"original": {
"owner": "mozilla",
"repo": "nixpkgs-mozilla",
"type": "github"
}
},
"naersk": {
"inputs": {
"nixpkgs": [
"nixpkgs"
]
},
"locked": {
"lastModified": 1614785451,
"narHash": "sha256-TPw8kQvr2UNCuvndtY+EjyXp6Q5GEW2l9UafXXh1XmI=",
"owner": "nmattia",
"repo": "naersk",
"rev": "e0fe990b478a66178a58c69cf53daec0478ca6f9",
"type": "github"
},
"original": {
"owner": "nmattia",
"ref": "master",
"repo": "naersk",
"type": "github"
}
},
"nixpkgs": {
"locked": {
"lastModified": 1616670887,
@ -88,8 +51,6 @@
"inputs": {
"flake-compat": "flake-compat",
"flake-utils": "flake-utils",
"mozillapkgs": "mozillapkgs",
"naersk": "naersk",
"nixpkgs": "nixpkgs"
}
}

View file

@ -1,13 +1,5 @@
{
inputs = {
naersk = {
url = "github:nmattia/naersk/master";
inputs.nixpkgs.follows = "nixpkgs";
};
mozillapkgs = {
url = "github:mozilla/nixpkgs-mozilla";
flake = false;
};
nixpkgs.url = "github:NixOS/nixpkgs/nixpkgs-unstable";
flake-utils.url = "github:numtide/flake-utils";
flake-compat = {
@ -16,27 +8,19 @@
};
};
outputs = { self, naersk, mozillapkgs, nixpkgs, flake-utils, ... }:
outputs = { self, nixpkgs, flake-utils, ... }:
flake-utils.lib.eachDefaultSystem (system:
let
pkgs = import nixpkgs { inherit system; };
mozilla = pkgs.callPackage (mozillapkgs + "/package-set.nix") { };
rustNightly = (mozilla.rustChannelOf {
date = "2021-03-29";
channel = "nightly";
sha256 = "sha256-Y94CnslybZgiZlNVV6Cg0TUPV2OeDXakPev1kqdt9Kk=";
}).rust;
naersk-lib = pkgs.callPackage naersk {
cargo = rustNightly;
rustc = rustNightly;
};
in
{
defaultPackage = naersk-lib.buildPackage {
src = ./.;
defaultPackage = pkgs.rustPlatform.buildRustPackage {
pname = "lohr";
version = "0.4.0";
src = ./.;
cargoSha256 = "sha256-5a2mK+E6LlR5RHDAhHDvnfPNG+0JdvpnL4kuTiz7vVg=";
meta = with pkgs.lib; {
description = "A Git mirroring tool";
@ -52,10 +36,12 @@
devShell = pkgs.mkShell {
buildInputs = with pkgs; [
cargo
clippy
nixpkgs-fmt
pre-commit
rustPackages.clippy
rustNightly
rustc
rustfmt
];

View file

@ -1 +1 @@
nightly
stable

View file

@ -4,7 +4,7 @@ use serde::Deserialize;
pub(crate) struct Repository {
pub(crate) name: String,
pub(crate) full_name: String,
pub(crate) clone_url: String,
pub(crate) ssh_url: String,
}
#[derive(Deserialize)]

View file

@ -38,7 +38,7 @@ impl Job {
let output = Command::new("git")
.arg("clone")
.arg("--mirror")
.arg(&self.repo.clone_url)
.arg(&self.repo.ssh_url)
.arg(format!("{}", self.local_path.as_ref().unwrap().display()))
.output()?;

View file

@ -1,5 +1,3 @@
#![feature(proc_macro_hygiene, decl_macro)]
use std::env;
use std::fs::File;
use std::path::{Path, PathBuf};
@ -10,7 +8,7 @@ use std::sync::{
use std::thread;
use anyhow::Context;
use clap::{App, Arg};
use clap::{crate_version, App, Arg};
use log::{error, info};
use rocket::{http::Status, post, routes, State};
@ -36,12 +34,15 @@ fn gitea_webhook(
config: State<GlobalSettings>,
) -> Status {
if config
.blacklist
.filters
.iter()
.any(|re| re.is_match(&payload.repository.full_name))
// Find first filter that matches the given destination
.find(|filter| filter.destination.as_ref().map_or(false, |re| re.is_match(&payload.repository.full_name)))
// Default to mirroring, unless told not to
.map_or(true, |filter| filter.mirror)
{
info!(
"Ignoring webhook for repo {} which is blacklisted",
"Ignoring webhook for repo {} which is marked as not mirrored",
payload.repository.full_name
);
return Status::Ok;
@ -92,9 +93,10 @@ fn parse_config(home: &Path, flags: &clap::ArgMatches) -> anyhow::Result<GlobalS
serde_yaml::from_reader(file).context("could not parse configuration file")
}
fn main() -> anyhow::Result<()> {
#[rocket::main]
async fn main() -> anyhow::Result<()> {
let matches = App::new("lohr")
.version("0.3.1")
.version(crate_version!())
.about("Git mirroring daemon")
.arg(
Arg::with_name("config")
@ -127,7 +129,8 @@ fn main() -> anyhow::Result<()> {
.manage(JobSender(Mutex::new(sender)))
.manage(Secret(secret))
.manage(config_state)
.launch();
.launch()
.await?;
Ok(())
}

View file

@ -10,8 +10,29 @@ pub(crate) struct GlobalSettings {
/// List of remote stems to use for every repository
#[serde(default)]
pub additional_remotes: Vec<RepoUrl>,
/// List of regexes, if a repository's name matches any of the, it is not mirrored by `lohr`
/// even if it contains a `.lorh` file.
#[serde(with = "serde_regex")]
pub blacklist: Vec<regex::Regex>,
/// List of filters to blacklist repositories, or modify push options on specific remotes.
/// Only the first matching filter is applied, so order is important.
#[serde(default)]
pub filters: Vec<FilterSettings>,
}
#[derive(Clone, Default, Deserialize)]
pub(crate) struct FilterSettings {
/// Match on the source remote
#[serde(with = "serde_regex", default)]
pub source: Option<regex::Regex>,
/// Match on the destination remote
#[serde(with = "serde_regex", default)]
pub destination: Option<regex::Regex>,
/// Whether to mirror the repository or not
#[serde(default = "default_true")]
pub mirror: bool,
/// Push options to be used for the matched remote
#[serde(default)]
pub push_options: Vec<String>,
}
// Workaround for https://github.com/serde-rs/serde/issues/368
fn default_true() -> bool {
true
}

View file

@ -1,20 +1,17 @@
use std::{
io::Read,
io,
ops::{Deref, DerefMut},
};
use rocket::{
data::{FromData, Outcome},
data::{ByteUnit, FromData, Outcome},
http::ContentType,
State,
};
use rocket::{
data::{Transform, Transformed},
http::Status,
};
use rocket::{http::Status, local_cache};
use rocket::{Data, Request};
use anyhow::anyhow;
use anyhow::{anyhow, Context};
use serde::Deserialize;
use crate::Secret;
@ -53,37 +50,29 @@ impl<T> DerefMut for SignedJson<T> {
}
}
const LIMIT: u64 = 1 << 20;
const LIMIT: ByteUnit = ByteUnit::Mebibyte(1);
impl<'r, T: Deserialize<'r>> SignedJson<T> {
fn from_str(s: &'r str) -> anyhow::Result<Self> {
serde_json::from_str(s)
.map(SignedJson)
.context("could not parse json")
}
}
// This is a one to one implementation of request_contrib::Json's FromData, but with HMAC
// validation.
//
// Tracking issue for chaining Data guards to avoid this:
// https://github.com/SergioBenitez/Rocket/issues/775
impl<'a, T> FromData<'a> for SignedJson<T>
#[rocket::async_trait]
impl<'r, T> FromData<'r> for SignedJson<T>
where
T: Deserialize<'a>,
T: Deserialize<'r>,
{
type Error = anyhow::Error;
type Owned = String;
type Borrowed = str;
fn transform(
request: &Request,
data: Data,
) -> rocket::data::Transform<Outcome<Self::Owned, Self::Error>> {
let size_limit = request.limits().get("json").unwrap_or(LIMIT);
let mut s = String::with_capacity(512);
match data.open().take(size_limit).read_to_string(&mut s) {
Ok(_) => Transform::Borrowed(Outcome::Success(s)),
Err(e) => Transform::Borrowed(Outcome::Failure((
Status::BadRequest,
anyhow!("couldn't read json: {}", e),
))),
}
}
fn from_data(request: &Request, o: Transformed<'a, Self>) -> Outcome<Self, Self::Error> {
async fn from_data(request: &'r Request<'_>, data: Data) -> Outcome<Self, Self::Error> {
let json_ct = ContentType::new("application", "json");
if request.content_type() != Some(&json_ct) {
return Outcome::Failure((Status::BadRequest, anyhow!("wrong content type")));
@ -97,26 +86,31 @@ where
));
}
let size_limit = request.limits().get("json").unwrap_or(LIMIT);
let content = match data.open(size_limit).into_string().await {
Ok(s) if s.is_complete() => s.into_inner(),
Ok(_) => {
let eof = io::ErrorKind::UnexpectedEof;
return Outcome::Failure((
Status::PayloadTooLarge,
io::Error::new(eof, "data limit exceeded").into(),
));
}
Err(e) => return Outcome::Failure((Status::BadRequest, e.into())),
};
let signature = signatures[0];
let secret = request.guard::<State<Secret>>().await.unwrap();
let content = o.borrowed()?;
let secret = request.guard::<State<Secret>>().unwrap();
if !validate_signature(&secret.0, &signature, content) {
if !validate_signature(&secret.0, &signature, &content) {
return Outcome::Failure((Status::BadRequest, anyhow!("couldn't verify signature")));
}
let content = match serde_json::from_str(content) {
Ok(content) => content,
Err(e) => {
return Outcome::Failure((
Status::BadRequest,
anyhow!("couldn't parse json: {}", e),
))
}
let content = match Self::from_str(local_cache!(request, content)) {
Ok(content) => Outcome::Success(content),
Err(e) => Outcome::Failure((Status::BadRequest, e)),
};
Outcome::Success(SignedJson(content))
content
}
}