microsoft/openvmm
Publicmirrored fromhttps://github.com/microsoft/openvmmAvailable
flowey/flowey_cli/src/pipeline_resolver/common_yaml.rs
269lines · modecode
| 1 | // Copyright (c) Microsoft Corporation. |
| 2 | // Licensed under the MIT License. |
| 3 | |
| 4 | //! Shared functionality for emitting a pipeline as ADO/GitHub YAML files |
| 5 | |
| 6 | use crate::cli::exec_snippet::FloweyPipelineStaticDb; |
| 7 | use crate::cli::pipeline::CheckMode; |
| 8 | use crate::pipeline_resolver::generic::ResolvedPipelineJob; |
| 9 | use anyhow::Context; |
| 10 | use flowey_core::node::FlowArch; |
| 11 | use flowey_core::node::FlowPlatform; |
| 12 | use petgraph::visit::EdgeRef; |
| 13 | use serde::Serialize; |
| 14 | use serde_yaml::Value; |
| 15 | use std::collections::BTreeMap; |
| 16 | use std::collections::BTreeSet; |
| 17 | use std::io::Write; |
| 18 | use std::path::Path; |
| 19 | |
| 20 | #[derive(Debug)] |
| 21 | pub(crate) enum FloweySource { |
| 22 | // bool indicates if this node should publish the flowey it bootstraps for |
| 23 | // other nodes to consume |
| 24 | Bootstrap(String, bool), |
| 25 | Consume(String), |
| 26 | } |
| 27 | |
| 28 | /// each job has one of three "roles" when it comes to bootstrapping flowey: |
| 29 | /// |
| 30 | /// 1. Build flowey |
| 31 | /// 2. Building _and_ publishing flowey |
| 32 | /// 3. Consuming a pre-built flowey |
| 33 | /// |
| 34 | /// We _could_ just have every bootstrap job also publish flowey, but this |
| 35 | /// will spam the artifact feed with artifacts no one will consume, which is |
| 36 | /// wasteful. |
| 37 | /// |
| 38 | /// META: why go through all this hassle anyways? i.e: why not just do |
| 39 | /// something dead simple like: |
| 40 | /// |
| 41 | /// - discover which platforms exist in the graph |
| 42 | /// - have the first jobs of every pipeline be standalone "bootstrap flowey" |
| 43 | /// jobs, which all subsequent jobs of a certain platform can take a dep on |
| 44 | /// |
| 45 | /// well... it turns out that provisioning job runners is _sloooooow_, |
| 46 | /// and having every single pipeline run these "bootstrap flowey" steps |
| 47 | /// gating the rest of the "interesting" stuff would really stink. |
| 48 | /// |
| 49 | /// i.e: it's better to do redundant flowey bootstraps if it means that we |
| 50 | /// can avoid the extra time it takes to tear down + re-provision a worker. |
| 51 | pub(crate) fn job_flowey_bootstrap_source( |
| 52 | graph: &petgraph::Graph<ResolvedPipelineJob, ()>, |
| 53 | order: &Vec<petgraph::prelude::NodeIndex>, |
| 54 | ) -> BTreeMap<petgraph::prelude::NodeIndex, FloweySource> { |
| 55 | let mut bootstrapped_flowey = BTreeMap::new(); |
| 56 | |
| 57 | // the first traversal builds a list of all ancestors of a give node |
| 58 | let mut ancestors = BTreeMap::< |
| 59 | petgraph::prelude::NodeIndex, |
| 60 | BTreeSet<(petgraph::prelude::NodeIndex, FlowPlatform, FlowArch)>, |
| 61 | >::new(); |
| 62 | for idx in order { |
| 63 | for ancestor_idx in graph |
| 64 | .edges_directed(*idx, petgraph::Direction::Incoming) |
| 65 | .map(|e| e.source()) |
| 66 | { |
| 67 | ancestors.entry(*idx).or_default().insert(( |
| 68 | ancestor_idx, |
| 69 | graph[ancestor_idx].platform, |
| 70 | graph[ancestor_idx].arch, |
| 71 | )); |
| 72 | |
| 73 | if let Some(set) = ancestors.get(&ancestor_idx).cloned() { |
| 74 | ancestors.get_mut(idx).unwrap().extend(&set); |
| 75 | } |
| 76 | } |
| 77 | } |
| 78 | |
| 79 | // the second traversal assigns roles to each node |
| 80 | let mut floweyno = 0; |
| 81 | 'outer: for idx in order { |
| 82 | let ancestors = ancestors.remove(idx).unwrap_or_default(); |
| 83 | |
| 84 | let mut elect_bootstrap = None; |
| 85 | |
| 86 | for (ancestor_idx, platform, arch) in ancestors { |
| 87 | if platform != graph[*idx].platform || arch != graph[*idx].arch { |
| 88 | continue; |
| 89 | } |
| 90 | |
| 91 | let role = |
| 92 | bootstrapped_flowey |
| 93 | .get_mut(&ancestor_idx) |
| 94 | .and_then(|existing| match existing { |
| 95 | FloweySource::Bootstrap(s, true) => Some(FloweySource::Consume(s.clone())), |
| 96 | FloweySource::Consume(s) => Some(FloweySource::Consume(s.clone())), |
| 97 | // there is an ancestor that is building, but not |
| 98 | // publishing. maybe they should get upgraded... |
| 99 | FloweySource::Bootstrap(_, false) => { |
| 100 | elect_bootstrap = Some(ancestor_idx); |
| 101 | None |
| 102 | } |
| 103 | }); |
| 104 | |
| 105 | if let Some(role) = role { |
| 106 | bootstrapped_flowey.insert(*idx, role); |
| 107 | continue 'outer; |
| 108 | } |
| 109 | } |
| 110 | |
| 111 | // if we got here, that means we couldn't find a valid ancestor. |
| 112 | // |
| 113 | // check if we can upgrade an existing ancestor vs. bootstrapping |
| 114 | // things ourselves |
| 115 | if let Some(elect_bootstrap) = elect_bootstrap { |
| 116 | let FloweySource::Bootstrap(s, publish) = |
| 117 | bootstrapped_flowey.get_mut(&elect_bootstrap).unwrap() |
| 118 | else { |
| 119 | unreachable!() |
| 120 | }; |
| 121 | |
| 122 | *publish = true; |
| 123 | let s = s.clone(); |
| 124 | |
| 125 | bootstrapped_flowey.insert(*idx, FloweySource::Consume(s)); |
| 126 | } else { |
| 127 | // Having this extra unique `floweyno` per bootstrap is |
| 128 | // necessary since GitHub doesn't let you double-publish an |
| 129 | // artifact with the same name |
| 130 | floweyno += 1; |
| 131 | let platform = graph[*idx].platform; |
| 132 | let arch = graph[*idx].arch; |
| 133 | bootstrapped_flowey.insert( |
| 134 | *idx, |
| 135 | FloweySource::Bootstrap( |
| 136 | format!("_internal-flowey-bootstrap-{arch}-{platform}-uid-{floweyno}"), |
| 137 | false, |
| 138 | ), |
| 139 | ); |
| 140 | } |
| 141 | } |
| 142 | |
| 143 | bootstrapped_flowey |
| 144 | } |
| 145 | |
| 146 | /// convert `pipeline` to YAML and `pipeline_static_db` to JSON. |
| 147 | /// if `check` is `Some`, then we will compare the generated YAML and JSON |
| 148 | /// against the contents of `check` and error if they don't match. |
| 149 | /// if `check` is `None`, then we will write the generated YAML and JSON to |
| 150 | /// `repo_root/pipeline_file.yaml` and `repo_root/pipeline_file.json` respectively. |
| 151 | fn check_or_write_generated_yaml_and_json<T>( |
| 152 | pipeline: &T, |
| 153 | pipeline_static_db: &FloweyPipelineStaticDb, |
| 154 | mode: CheckMode, |
| 155 | repo_root: &Path, |
| 156 | pipeline_file: &Path, |
| 157 | ado_post_process_yaml_cb: Option<Box<dyn FnOnce(Value) -> Value>>, |
| 158 | ) -> anyhow::Result<()> |
| 159 | where |
| 160 | T: Serialize, |
| 161 | { |
| 162 | let generated_yaml = |
| 163 | serde_yaml::to_value(pipeline).context("while serializing pipeline yaml")?; |
| 164 | let generated_yaml = if let Some(ado_post_process_yaml_cb) = ado_post_process_yaml_cb { |
| 165 | ado_post_process_yaml_cb(generated_yaml) |
| 166 | } else { |
| 167 | generated_yaml |
| 168 | }; |
| 169 | |
| 170 | let generated_yaml = |
| 171 | serde_yaml::to_string(&generated_yaml).context("while emitting pipeline yaml")?; |
| 172 | let generated_yaml = format!( |
| 173 | r#" |
| 174 | ############################## |
| 175 | # THIS FILE IS AUTOGENERATED # |
| 176 | # DO NOT MANUALLY EDIT # |
| 177 | ############################## |
| 178 | {generated_yaml}"# |
| 179 | ); |
| 180 | let generated_yaml = generated_yaml.trim_start(); |
| 181 | |
| 182 | let generated_json = |
| 183 | serde_json::to_string_pretty(pipeline_static_db).context("while emitting pipeline json")?; |
| 184 | |
| 185 | match mode { |
| 186 | CheckMode::Runtime(ref check_file) | CheckMode::Check(ref check_file) => { |
| 187 | let existing_yaml = fs_err::read_to_string(check_file) |
| 188 | .context("cannot check pipeline that doesn't exist!")?; |
| 189 | |
| 190 | let yaml_out_of_date = existing_yaml != generated_yaml; |
| 191 | |
| 192 | if yaml_out_of_date { |
| 193 | println!( |
| 194 | "generated yaml {}:\n==========\n{generated_yaml}", |
| 195 | generated_yaml.len() |
| 196 | ); |
| 197 | println!( |
| 198 | "existing yaml {}:\n==========\n{existing_yaml}", |
| 199 | existing_yaml.len() |
| 200 | ); |
| 201 | } |
| 202 | |
| 203 | if yaml_out_of_date { |
| 204 | anyhow::bail!("checked in pipeline YAML is out of date! run `cargo xflowey regen`") |
| 205 | } |
| 206 | |
| 207 | // Only write the JSON if we're in runtime mode, not in check mode |
| 208 | if let CheckMode::Runtime(_) = mode { |
| 209 | let mut f = fs_err::File::create(check_file.with_extension("json"))?; |
| 210 | f.write_all(generated_json.as_bytes()) |
| 211 | .context("while emitting pipeline database json")?; |
| 212 | } |
| 213 | |
| 214 | Ok(()) |
| 215 | } |
| 216 | CheckMode::None => { |
| 217 | let out_yaml_path = repo_root.join(pipeline_file); |
| 218 | |
| 219 | let mut f = fs_err::File::create(out_yaml_path)?; |
| 220 | f.write_all(generated_yaml.as_bytes()) |
| 221 | .context("while emitting pipeline yaml")?; |
| 222 | |
| 223 | Ok(()) |
| 224 | } |
| 225 | } |
| 226 | } |
| 227 | |
| 228 | /// See [`check_or_write_generated_yaml_and_json`] |
| 229 | pub(crate) fn check_generated_yaml_and_json<T>( |
| 230 | pipeline: &T, |
| 231 | pipeline_static_db: &FloweyPipelineStaticDb, |
| 232 | check: CheckMode, |
| 233 | repo_root: &Path, |
| 234 | pipeline_file: &Path, |
| 235 | ado_post_process_yaml_cb: Option<Box<dyn FnOnce(Value) -> Value>>, |
| 236 | ) -> anyhow::Result<()> |
| 237 | where |
| 238 | T: Serialize, |
| 239 | { |
| 240 | check_or_write_generated_yaml_and_json( |
| 241 | pipeline, |
| 242 | pipeline_static_db, |
| 243 | check, |
| 244 | repo_root, |
| 245 | pipeline_file, |
| 246 | ado_post_process_yaml_cb, |
| 247 | ) |
| 248 | } |
| 249 | |
| 250 | /// See [`check_or_write_generated_yaml_and_json`] |
| 251 | pub(crate) fn write_generated_yaml_and_json<T>( |
| 252 | pipeline: &T, |
| 253 | pipeline_static_db: &FloweyPipelineStaticDb, |
| 254 | repo_root: &Path, |
| 255 | pipeline_file: &Path, |
| 256 | ado_post_process_yaml_cb: Option<Box<dyn FnOnce(Value) -> Value>>, |
| 257 | ) -> anyhow::Result<()> |
| 258 | where |
| 259 | T: Serialize, |
| 260 | { |
| 261 | check_or_write_generated_yaml_and_json( |
| 262 | pipeline, |
| 263 | pipeline_static_db, |
| 264 | CheckMode::None, |
| 265 | repo_root, |
| 266 | pipeline_file, |
| 267 | ado_post_process_yaml_cb, |
| 268 | ) |
| 269 | } |
| 270 | |