Fix data corruption (shard.go) #119
Conversation
|
I pushed a commit on top which fixes a datarace in |
|
Before the fix in With the last fix, it passes Since I now do the read twice in |
holiman
changed the title
| @@ -162,6 +162,11 @@ func (q *BytesQueue) Get(index int) ([]byte, error) { | |||
| return data, err | |||
| } | |||
|
|
|||
| // CheckGet checks if an entry can be read from index | |||
| func (q *BytesQueue) CheckGet(index int) error { | |||
There was a problem hiding this comment.
Please add a test for this.
| @@ -177,6 +182,23 @@ func (e *queueError) Error() string { | |||
| return e.message | |||
| } | |||
|
|
|||
| // peekCheckErr is identical to peek, but does not actually return any data | |||
| func (q *BytesQueue) peekCheckErr(index int) error { | |||
There was a problem hiding this comment.
Also needs a test. :)
|
I addressed the concerns. |
|
So that's weird. I merged it in the UI, it didn't merge the branch, and then it just kicked off another travis build, but then travis errored out: $HOME/gopath/bin/goveralls -coverprofile=gover.coverprofile -service travis-ci
Bad response status from coveralls: 422
{"message":"Couldn't find a repository matching this job.","error":true}
The command "$HOME/gopath/bin/goveralls -coverprofile=gover.coverprofile -service travis-ci" exited with 1.@cristaloleg can you take a look when you get a chance? |
|
Thank you @holiman |
This PR contains a testcase which demonstrates corruption (intermittently). Sometimes leading to `panic`, and sometimes leading to data corruption of values. I thought that fixing the datarace suggested in allegro#117 would solve it, but it seems I was wrong about that, there's some other underlying bug. I'll push a commit on top if I find it.
This PR contains a testcase which demonstrates corruption (intermittently). Sometimes leading to `panic`, and sometimes leading to data corruption of values. I thought that fixing the datarace suggested in allegro#117 would solve it, but it seems I was wrong about that, there's some other underlying bug. I'll push a commit on top if I find it.
This PR contains a testcase which demonstrates corruption (intermittently). Sometimes leading to
panic, and sometimes leading to data corruption of values.I thought that fixing the datarace suggested in #117 would solve it, but it seems I was wrong about that, there's some other underlying bug. I'll push a commit on top if I find it.