Bleve indexer stuck on large repo

[Fogapod]

Description

I'm running gitea on weak hardware with just 1 cpu. I got linux kernel mirror to work but code indexing is in a weird state.
Here is what admin ui shows:

All other repos have commit shown so indexer works otherwise.

When i add reindexing to queue, multiple processes appear in system and run for a couple minutes, then they disappear. There are no dmesg OOM messages, there are no gitea logs about it. Gitea admin ui still shows them as running:

Code search works, at least partially:

I have nothing interesting in config:

[indexer]
REPO_INDEXER_ENABLED = true

gitea.com has linux source. I wonder what is the state of indexer there? https://gitea.com/marktsai0316/linux
gitea.com has indexer disabled

Gitea Version

1.17.0+dev-751-g433443ffa built with GNU Make 4.3, go1.18.1 : bindata

Can you reproduce the bug on the Gitea demo site?

No

Log Gist

No response

Screenshots

No response

Git Version

2.34.1

Operating System

Ubuntu 22.04 LTS

How are you running Gitea?

Locally compiled binary

Database

PostgreSQL

[Fogapod]

Clicking trash icon does nothing:

Jun 18 17:04:15 localhost gitea[62335]: 2022/06/18 17:04:15 [62ae058f] router: completed POST /admin/monitor/cancel/62adf91d-2 for [::]:0, 200 OK in 3.9ms @ admin/admin.go:358(admin.MonitorCancel)
Jun 18 17:04:15 localhost gitea[62335]: 2022/06/18 17:04:15 [62ae056b] router: completed GET /user/events for [::]:0, 200 OK in 36032.1ms @ events/events.go:19(events.Events)
Jun 18 17:04:15 localhost gitea[62335]: 2022/06/18 17:04:15 [62ae058f-2] router: completed GET /admin/monitor for [::]:0, 200 OK in 25.1ms @ admin/admin.go:326(admin.Monitor)
Jun 18 17:04:19 localhost gitea[62335]: 2022/06/18 17:04:19 [62adf4fd] router: polling   GET /user/events for [::]:0, elapsed 3439.0ms @ events/events.go:19(events.Events)

[wxiaoguang]

It's highly likely related to this problem, but there is no clue about how to fix it.

gitea/modules/indexer/code/indexer.go

Lines 146 to 163 in c090f87

	for _, indexerData := range items {
	log.Trace("IndexerData Process Repo: %d", indexerData.RepoID)
	// FIXME: it seems there is a bug in `CatFileBatch` or `nio.Pipe`, which will cause the process to hang forever in rare cases
	/*
	sync.(*Cond).Wait(cond.go:70)
	github.com/djherbis/nio/v3.(*PipeReader).Read(sync.go:106)
	bufio.(*Reader).fill(bufio.go:106)
	bufio.(*Reader).ReadSlice(bufio.go:372)
	bufio.(*Reader).collectFragments(bufio.go:447)
	bufio.(*Reader).ReadString(bufio.go:494)
	code.gitea.io/gitea/modules/git.ReadBatchLine(batch_reader.go:149)
	code.gitea.io/gitea/modules/indexer/code.(*BleveIndexer).addUpdate(bleve.go:214)
	code.gitea.io/gitea/modules/indexer/code.(*BleveIndexer).Index(bleve.go:296)
	code.gitea.io/gitea/modules/indexer/code.(*wrappedIndexer).Index(wrapped.go:74)
	code.gitea.io/gitea/modules/indexer/code.index(indexer.go:105)
	*/
	if err := index(ctx, indexer, indexerData.RepoID); err != nil {

[wxiaoguang]

It should have been fixed, related to #26325, more details in #26967