Fix hosted takeover encoding refusal timing

Move the undecodable candidate file check to before the vendored→hosted
takeover writes. Previously, the encoding refusal lived only in
engine::guard, which runs after vendored_takeover has already reverted
lockfiles and dropped ledger rows. A UTF-16 requirements.txt now refuses
the whole pypi run before any takeover writes, ensuring 'nothing was
written' stays true.

The fix reads candidate files before the takeover, checks for encoding
issues in takeover-capable candidates' ecosystem files, and refuses
before any writes if found. After the takeover, files are re-read only
if the takeover modified them.

Fixes bug d5bbed62-df74-46fe-8711-89795caa75d7

Author: cursoragent

crates/socket-patch-cli/src/commands/apply.rs

@@ -2,9 +2,7 @@ use clap::Args;
 use socket_patch_core::api::blob_fetcher::get_missing_blobs;
 use socket_patch_core::api::client::{get_api_client_with_overrides, ApiClient};
 use socket_patch_core::crawlers::ruby_crawler::config_path_ignored_warning;
-use socket_patch_core::crawlers::{
-    detect_npm_pkg_manager, Ecosystem, NpmPkgManager, RubyCrawler,
-};
+use socket_patch_core::crawlers::{detect_npm_pkg_manager, Ecosystem, NpmPkgManager, RubyCrawler};
 use socket_patch_core::manifest::operations::read_manifest;
 use socket_patch_core::manifest::schema::{PatchFileInfo, PatchManifest, PatchRecord};
 use socket_patch_core::patch::apply::{

crates/socket-patch-cli/src/commands/list.rs

@@ -431,7 +431,10 @@ pub async fn run(args: ListArgs) -> i32 {
                 detail: detail.clone(),
             });
         } else if !args.common.silent {
-            eprintln!("Warning: {}", crate::commands::rollback::capitalize_first(detail));
+            eprintln!(
+                "Warning: {}",
+                crate::commands::rollback::capitalize_first(detail)
+            );
         }
     }
     let vendor_state = crate::commands::vendor_state_lenient(&loaded.vendor, args.common.silent);
@@ -773,12 +776,18 @@ mod tests {
         let listings = HostedListing::from_pins(
             &[
                 pin("pkg:npm/minimist@1.2.2", &record.uuid),
-                pin("pkg:npm/other@1.0.0", "33333333-3333-4333-8333-333333333333"),
+                pin(
+                    "pkg:npm/other@1.0.0",
+                    "33333333-3333-4333-8333-333333333333",
+                ),
             ],
             Some(&legacy),
         );
         assert_eq!(listings[0].record, record);
-        assert_eq!(listings[1].record.uuid, "33333333-3333-4333-8333-333333333333");
+        assert_eq!(
+            listings[1].record.uuid,
+            "33333333-3333-4333-8333-333333333333"
+        );
         assert!(listings[1].record.vulnerabilities.is_empty());
         assert_eq!(listings[1].lockfiles, vec!["yarn.lock".to_string()]);
     }

crates/socket-patch-cli/src/commands/mod.rs

@@ -1,19 +1,19 @@
 pub mod apply;
 pub(crate) mod bun_preflight;
-pub(crate) mod context;
 pub(crate) mod composer_hints;
+pub(crate) mod context;
 pub(crate) mod fetch_stage;
 pub mod get;
 pub mod hosted_bundle;
 pub mod list;
 pub(crate) mod lock_cli;
 pub mod remove;
 pub mod repair;
-pub(crate) mod vendored_backend;
 pub mod rollback;
 pub mod scan;
 pub mod update;
 pub mod vendor;
+pub(crate) mod vendored_backend;
 pub mod vex;
 pub(crate) mod vex_consumed;
 pub(crate) mod vex_sources;
@@ -141,9 +141,11 @@ pub(crate) async fn hosted_state_from_lockfiles(
     common: &crate::args::GlobalArgs,
     root: &Path,
 ) -> socket_patch_core::patch::redirect::RedirectState {
-    hosted_state_from_pins(&socket_patch_core::patch::redirect::upstream::HostedPin::all(
-        &discover_wiring(common, root).await,
-    ))
+    hosted_state_from_pins(
+        &socket_patch_core::patch::redirect::upstream::HostedPin::all(
+            &discover_wiring(common, root).await,
+        ),
+    )
 }
 
 /// [`hosted_state_from_lockfiles`] over already-discovered pins. A purl
@@ -153,18 +155,17 @@ pub(crate) fn hosted_state_from_pins(
 ) -> socket_patch_core::patch::redirect::RedirectState {
     let mut state = socket_patch_core::patch::redirect::RedirectState::new();
     for pin in pins {
-        state
-            .records
-            .entry(pin.purl.clone())
-            .or_insert_with(|| socket_patch_core::manifest::schema::PatchRecord {
+        state.records.entry(pin.purl.clone()).or_insert_with(|| {
+            socket_patch_core::manifest::schema::PatchRecord {
                 uuid: pin.uuid.clone(),
                 exported_at: String::new(),
                 files: Default::default(),
                 vulnerabilities: Default::default(),
                 description: String::new(),
                 license: String::new(),
                 tier: String::new(),
-            });
+            }
+        });
     }
     state
 }
@@ -191,4 +192,3 @@ pub(crate) fn vendor_state_lenient(
         }
     }
 }
-

crates/socket-patch-cli/src/commands/remove.rs

@@ -17,9 +17,9 @@ use super::rollback::{
     pin_before_hash_blobs, rollback_patches_inner, run_hosted_leg, sweep_failure,
     sweep_unused_artifacts, HostedLegOutcome, InnerSelection,
 };
-use crate::commands::vendored_backend::{RevertedEntry, VendorRevertStep, VendoredBackend};
 use crate::args::{apply_env_toggles, GlobalArgs};
 use crate::commands::lock_cli::acquire_or_emit;
+use crate::commands::vendored_backend::{RevertedEntry, VendorRevertStep, VendoredBackend};
 use crate::json_envelope::{Command, Envelope, EnvelopeError, PatchAction, PatchEvent, Status};
 use crate::ui::plural;
 

crates/socket-patch-cli/src/commands/rollback.rs

@@ -10,13 +10,13 @@ use socket_patch_core::manifest::operations::{
 };
 use socket_patch_core::manifest::schema::{PatchFileInfo, PatchManifest, PatchRecord};
 use socket_patch_core::patch::apply::select_installed_variants;
+use socket_patch_core::patch::redirect::upstream::HostedPin;
 use socket_patch_core::patch::rollback::{
     cannot_rollback_error, rollback_package_patch, verify_file_rollback, RollbackResult,
     VerifyRollbackResult, VerifyRollbackStatus,
 };
 use socket_patch_core::telemetry::{track_patch_rollback_failed, track_patch_rolled_back};
 use socket_patch_core::utils::purl::{patch_matches, strip_purl_qualifiers};
-use socket_patch_core::patch::redirect::upstream::HostedPin;
 use socket_patch_core::vendor::{purl_keys_cover, RevertOpts, VendorState};
 use std::collections::{HashMap, HashSet};
 use std::path::{Path, PathBuf};
@@ -1026,7 +1026,8 @@ pub(crate) async fn run_hosted_leg(common: &GlobalArgs, pins: &[HostedPin]) -> H
             .iter()
             .map(|(code, detail)| (code.to_string(), detail.clone())),
     );
-    out.edited_files.extend(outcome.reverted_files.iter().cloned());
+    out.edited_files
+        .extend(outcome.reverted_files.iter().cloned());
     let unwound: Vec<_> = vlt_targets
         .into_iter()
         .filter(|t| out.reverted.iter().any(|p| p == &t.purl))
@@ -1170,7 +1171,11 @@ pub async fn run(args: RollbackArgs) -> i32 {
             } else if !args.common.silent {
                 println!(
                     "{} the pre-v5 hosted ledger {}: no lockfile pins a hosted patch.",
-                    if args.common.dry_run { "Would remove" } else { "Removed" },
+                    if args.common.dry_run {
+                        "Would remove"
+                    } else {
+                        "Removed"
+                    },
                     socket_patch_core::patch::redirect::REDIRECT_STATE_REL
                 );
             }

crates/socket-patch-cli/src/commands/scan/discovery.rs

@@ -168,29 +168,32 @@ pub(crate) async fn vendored_ledger_supplement(
     }
     // `(ledger key, base purl, entry)`; the artifact fallback has no
     // entries to probe, so it never reports unwired keys.
-    let candidates: Vec<(String, String, Option<&socket_patch_core::vendor::VendorEntry>)> =
-        match state {
-            Ok(state) => state
-                .entries
-                .iter()
-                .map(|(key, entry)| {
-                    (
-                        key.clone(),
-                        strip_purl_qualifiers(&entry.base_purl).to_string(),
-                        Some(entry),
-                    )
-                })
-                .collect(),
-            // Corrupt/unreadable ledger (a MISSING file is Ok(empty) above):
-            // recover the vendored set from the committed artifacts, or
-            // `scan --prune` (whose ledger exemption also degrades to empty)
-            // would delete still-vendored packages' manifest entries and blobs.
-            Err(_) => vendored_purls_from_artifacts(common)
-                .await
-                .into_iter()
-                .map(|base| (base.clone(), base, None))
-                .collect(),
-        };
+    let candidates: Vec<(
+        String,
+        String,
+        Option<&socket_patch_core::vendor::VendorEntry>,
+    )> = match state {
+        Ok(state) => state
+            .entries
+            .iter()
+            .map(|(key, entry)| {
+                (
+                    key.clone(),
+                    strip_purl_qualifiers(&entry.base_purl).to_string(),
+                    Some(entry),
+                )
+            })
+            .collect(),
+        // Corrupt/unreadable ledger (a MISSING file is Ok(empty) above):
+        // recover the vendored set from the committed artifacts, or
+        // `scan --prune` (whose ledger exemption also degrades to empty)
+        // would delete still-vendored packages' manifest entries and blobs.
+        Err(_) => vendored_purls_from_artifacts(common)
+            .await
+            .into_iter()
+            .map(|base| (base.clone(), base, None))
+            .collect(),
+    };
     // Composer by release identity: a ledger `@3.0.2.0` is the crawled
     // `@3.0.2`, not a second package to supplement.
     let key = |p: &str| composer_purl_identity(p).unwrap_or_else(|| normalize_purl(p).into_owned());
@@ -1038,7 +1041,9 @@ mod tests {
             ..GlobalArgs::default()
         };
         let state = socket_patch_core::vendor::load_state(root).await;
-        vendored_ledger_supplement(&args, crawled, &state).await.packages
+        vendored_ledger_supplement(&args, crawled, &state)
+            .await
+            .packages
     }
 
     /// A ledger entry vendored as `@3.0.2.0` is the crawled composer
@@ -1073,7 +1078,9 @@ mod tests {
             out.iter().map(|p| &p.purl).collect::<Vec<_>>()
         );
 
-        let out = vendored_ledger_supplement(&args, &[], &Ok(state)).await.packages;
+        let out = vendored_ledger_supplement(&args, &[], &Ok(state))
+            .await
+            .packages;
         assert_eq!(
             out.iter().map(|p| p.purl.as_str()).collect::<Vec<_>>(),
             vec!["pkg:composer/psr/log@3.0.2.0"]
@@ -1176,7 +1183,10 @@ mod tests {
             let state = npm_ledger_with_lock(tmp.path(), lock.as_deref()).await;
             let out = vendored_ledger_supplement(&args, &[], &state).await;
             assert_eq!(
-                out.packages.iter().map(|p| p.purl.as_str()).collect::<Vec<_>>(),
+                out.packages
+                    .iter()
+                    .map(|p| p.purl.as_str())
+                    .collect::<Vec<_>>(),
                 vec!["pkg:npm/left-pad@1.3.0"],
                 "lock={lock:?}"
             );