When a db write fails because MaxSize is reached, the db client should still accept future reads and writes

[mattsains]

I found a bug with my work to introduce a maximum database size feature (#929).

If a write fails because the maximum size of the database is exceeded, it correctly returns an error explaining the constraint.

This is tested here: https://github.com/etcd-io/bbolt/blob/main/db_test.go#L1513

However, it seems to me that if a write fails for this reason, it should not "invalidate" the entire db client and future reads and writes should still work as long as the size is not exceeded again.

Instead, the behaviour is that the database client gets into a bad state because of the incomplete mmap operation. Before returning the MaxSizeReached error, we should re-mmap the db to the original size to allow continued operations

To confirm that this bug is fixed, the above test should perform a final, small write which should be verified to succeed.

[mattsains]

/assign-me

[ahrtr]

I think we need to re-implement the file limitation feature. #929

We just need to control the max size in only one place, between the two lines below, before mmap,

bbolt/db.go

Lines 1168 to 1169 in cbc4e1f

	if minsz >= db.datasz {
	if err := db.mmap(minsz); err != nil {

[ahrtr]

/assign @mattsains

thx for raising this.