golem15/msd-core
3
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix(#3610): hoist surviving top-level codex config keys to file scope on merge (#3690)

Browse Source 
* test(#3610): pin top-level key hoisting above the codex managed block

* fix(#3610): hoist surviving top-level keys above the codex managed block

* chore(#3610): add changeset

* fix(#3610): hoist to file scope (before the first table header) with reviewer-driven coverage

* chore(#3610): backfill changeset pr number

---------

Co-authored-by: sim <sim@local>
This commit is contained in:
Tom Boucher
2026-08-19 18:38:13 -04:00
committed by GitHub
parent 8526bd46f8
commit bad1f045b1
3 changed files with 179 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

44
bin/install.js
View File

@@ -6206,6 +6206,31 @@ const __atomicWrittenTmps = hooksSurface.__atomicWrittenTmps;
* All writes go through atomicWriteFileSync so a mid-write failure leaves
* the original config.toml untouched (#2760 fix 4).
*/
/**
* Split TOML content into its leading TOP-LEVEL key lines and everything from
* the first table header onward (#3610).
*
* Top-level keys were file-scoped before a merge. The regenerated GSD block
* opens with a table header (`[agents]`, #2088/ADR-1239 upgrade 2), so placing
* that block ABOVE surviving top-level keys re-scopes them into `[agents]` —
* `validateCodexConfigSchema` then correctly rejects the merged file and the
* install aborts. Hoisting the keys above the block preserves their scope.
*
* Table headers inside multiline strings do not start the "rest" region (the
* record parser already excludes them via startsInMultilineString).
*/
function splitTopLevelKeys(content) {
for (const record of getTomlLineRecords(content)) {
if (record.tableHeader && !record.startsInMultilineString) {
return {
topLevel: content.slice(0, record.start).trim(),
rest: content.slice(record.start).trim(),
};
}
}
return { topLevel: content.trim(), rest: '' };
}
function mergeCodexConfig(configPath, gsdBlock) {
// Case 1: No config.toml — create fresh
if (!fs.existsSync(configPath)) {
@@ -6253,10 +6278,25 @@ function mergeCodexConfig(configPath, gsdBlock) {
.replace(/^\r?\n# GSD codex_hooks ownership: (?:section|root_dotted)\r?\n/, '');
const afterUser = stripLeakedGsdCodexSections(markerStripped).trim();
// #3610: top-level keys that survived BELOW the marker were file-scoped
// before this merge; the regenerated block opens with the `[agents]` table
// header, so they must be hoisted to FILE scope or TOML re-scopes them
// into a table. File scope means BEFORE the first table header of the
// pre-marker region too — appending them after a pre-marker table (the
// default real-world layout: user tables precede the marker) would merely
// capture them into THAT table instead of [agents], and the schema
// validator is blind to non-agents tables.
const beforeSplit = before ? splitTopLevelKeys(before) : { topLevel: '', rest: '' };
const { topLevel: afterTopLevel, rest: afterTables } = splitTopLevelKeys(afterUser);
const parts = [];
if (before) parts.push(before);
const topParts = [];
if (beforeSplit.topLevel) topParts.push(beforeSplit.topLevel);
if (afterTopLevel) topParts.push(afterTopLevel);
if (topParts.length > 0) parts.push(topParts.join(eol + eol));
if (beforeSplit.rest) parts.push(beforeSplit.rest);
parts.push(normalizedGsdBlock);
if (afterUser) parts.push(afterUser);
if (afterTables) parts.push(afterTables);
atomicWriteFileSync(configPath, parts.join(eol + eol) + eol);
return;
}