fix: panic when update and iterate simultaneously

[WideLee]

1. Fix panic when update and iteration simultaneously. Related to Iterator readEntry crashes #222.
2. Add panic recover in cleanUp to prevent the program exited. Related to panic: runtime error: index out of range [7] with length 1 #226, panic out of range in bytes queue #148, just protect the main program not exited.
3. Byte queue set full is false after allocated addition memory. Also added a test case reproduce this problem.

[janisz]

Why do we need this?

[WideLee]

In #226 panic in cleanUp.

If CleanWindows is not zero, bigcache starts a goroutine, if cleanUp panic, the whole program will be exited.

bigcache/bigcache.go

Lines 85 to 99 in bbbffd3

	if config.CleanWindow > 0 {
	go func() {
	ticker := time.NewTicker(config.CleanWindow)
	defer ticker.Stop()
	for {
	select {
	case t := <-ticker.C:
	cache.cleanUp(uint64(t.Unix()))
	case <-cache.close:
	return
	}
	}
	}()
	}

This is a temporary solution to prevent crash of whole program, actually, find out the reason why panic is important.

[siennathesane]

I think it looks good to me, it could just use adjustments based off @janisz's comments.

[WideLee]

I think it looks good to me, it could just use adjustments based off @janisz's comments.

Do you mean that I should rollback the commit about panic recover in CleanUp?

[WideLee]

@janisz @mxplusb I just reverted the panic recover commit, please review it again 😄

[siennathesane]

It looks okay to me. I wanted to take the panic recovery out because I want to see how much it's happening across the board and fix the root of the problem when it becomes a much larger issue.

[exobin]

Is there any progress on solving this underlying issue? We are still getting these crashes 🤦

[siennathesane]

@alexi you're still having this same issue?

[exobin]

@mxplusb now getting this error:

runtime error: slice bounds out of range [:46697172] with capacity 585000
goroutine [running]:
       runtime/panic.go:969 +0x166
github.com/allegro/bigcache/queue.(*BytesQueue).peek(0xc06f6d0008, 0x22eec, 0x145676292821921e, 0xc06f6cb148, 0xc0be8f2600, 0xc0be8f2720, 0x2, 0x0)
        github.com/allegro/bigcache/queue/bytes_queue.go:240 +0x17c
github.com/allegro/bigcache/queue.(*BytesQueue).Get(...)
        github.com/allegro/bigcache/queue/bytes_queue.go:191
github.com/allegro/bigcache.(*cacheShard).getWrappedEntry(0xc06f6d0000, 0x145676292821921e, 0x0, 0xc127e1cbf8, 0x407a09, 0x0, 0xc0be8f25d0)
       github.com/allegro/bigcache/shard.go:110 +0x6e
github.com/allegro/bigcache.(*cacheShard).get(0xc06f6d0000, 0xc1280310c8, 0x8, 0x145676292821921e, 0xc167cc8840, 0x6bd0dc28, 0xc0c0b19078, 0x149c2f0, 0xc1492bdbd0)
       github.com/allegro/bigcache/shard.go:64 +0x64
github.com/allegro/bigcache.(*BigCache).Get(0xc000933110, 0xc1280310c8, 0x8, 0xf, 0x1, 0xc0001de3a0, 0xf, 0xf)
       github.com/allegro/bigcache/bigcache.go:117 +0x8b

[siennathesane]

@alexi can you please verify you are using this version or latest?

[exobin]

Yes, using commit 7bf29c0 (most recent).

The issues section has multiple examples of this bug: #226, and #148 has an open pull request fixing a uint32 overflow which looks like the likely culprit.

[siennathesane]

@alexi yeah, I think #148 could fix this.

[exobin]

@mxplusb Any idea when it might be merged?

[siennathesane]

No ETA, we're prepping for v3, which should include a fix for this. I'll try and get some time spent on this soon.

[fpessolano]

Any news, we have now the same issues on production servers (which we let crash and restart at the moment)

[exobin]

Hey @mxplusb any update here? We are in the same boat as @fpessolano. If it's possible to deploy it as a smaller patch that would be very valuable.

[siennathesane]

We're prepping for v3 in the versions/v3 branch, which should include the fix for this, but no current ETA on the release. I might be able to spend some time on it this month.