fix(confgen): preserve sub-table book/sheet name in merger errors

[Kybxd]

Summary

When a merger sub-table fails to parse, error messages incorrectly attributed the failure to the primary book/sheet instead of the offending sub-table. This PR fixes the attribution and adds an integration test to lock the behavior in.

Problem

ParseMessage runs parseMessageFromOneImporter for each merger importer in a goroutine. When one fails, the returned error was wrapped only with PrimaryBookName / PrimarySheetName:

return xerrors.WrapKV(err,
    xerrors.KeyPrimaryBookName, info.PrimaryBookName,
    xerrors.KeyPrimarySheetName, info.SheetOpts.Name)

Collector retains only the outermost withMessage as outerWM, so this outer WrapKV overwrote the inner layer that carried the actual sub-table's BookName / SheetName. The final NewDesc therefore pointed at the primary table rather than the shard that actually failed.

Fix

Attach the sub-table's BookName / SheetName on the same outer WrapKV layer alongside the primary names, so all four fields survive to NewDesc:

bookName := getRelBookName(info.ExtInfo.InputDir, impInfo.Filename())
sheetName := getRealSheetName(info, impInfo)
protomsg, err := parseMessageFromOneImporter(info, collector, impInfo)
if err != nil {
    return xerrors.WrapKV(err,
        xerrors.KeyBookName, bookName,
        xerrors.KeySheetName, sheetName,
        xerrors.KeyPrimaryBookName, info.PrimaryBookName,
        xerrors.KeyPrimarySheetName, info.SheetOpts.Name)
}

bookName / sheetName are now computed once and reused for both the error path and the success-path oneMsg.

Test

Added internal/confgen/testdata/collector_merger/ containing:

• proto/merger.proto — a workbook (MergerCollector#*.csv) with one sheet whose merger option pulls MergerShard*.csv#MergerCollectorItemConf.
• csv/merger/MergerCollector#@TABLEAU.csv, MergerCollector#MergerCollectorItemConf.csv, MergerShard1#MergerCollectorItemConf.csv — fixtures triggering a parse error in the shard.

collector_integration_test.go asserts the error description contains the shard book/sheet (MergerShard1 / MergerCollectorItemConf) rather than only the primary book/sheet.

Why a separate top-level testdata dir?

buildWorkbookIndex resolves merger shard globs in the test's inputDir. If merger.proto lived under the shared collector/proto/ directory, the normal/ and overflow/ test suites — which reuse the same collectortest proto package — would fail to resolve MergerShard*.csv in their own inputDirs. An isolated collector_merger/ dir with its own proto package avoids polluting those suites; the trade-off is one extra top-level testdata directory.

Risk

Low. Pure additive on the error path (extra KV fields) plus new test fixtures. No behavior change on the success path beyond hoisting two local variables.

[github-actions]

The latest Buf updates on your PR. Results from workflow Buf CI / buf (pull_request).

Build	Format	Lint	Breaking	Updated (UTC)
✅ passed	✅ passed	✅ passed	✅ passed	Jun 18, 2026, 7:46 AM

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 74.94%. Comparing base (1743f92) to head (9a93349).

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #420      +/-   ##
==========================================
+ Coverage   74.87%   74.94%   +0.07%     
==========================================
  Files          88       88              
  Lines        9384     9388       +4     
==========================================
+ Hits         7026     7036      +10     
+ Misses       1785     1781       -4     
+ Partials      573      571       -2     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.