Channel creation leaks heap memory since 7c59a8fa31c3 #1676
Description
Before filing a bug, please check whether it has been fixed since
the latest release: run "hg pull -u" and retry what you did to
reproduce the problem. Thanks.
What steps will reproduce the problem?
1. compile this code
package main
import (
"runtime"
"fmt"
"time"
)
func stats() {
for {
<-time.After(3e9)
fmt.Printf("CurrAlloc: %d, HeapAlloc: %d\n", runtime.MemStats.Alloc, runtime.MemStats.HeapAlloc)
}
}
func main() {
go stats()
for {
<-time.After(7e9)
runtime.GC()
}
}
2. run and wait (10 mins is sufficient) 6g leaky.go ; 6l leaky.6 ; ./6.out
What is the expected output?
CurrAlloc should stabalise, eg
CurrAlloc: 345224, HeapAlloc: 344784
CurrAlloc: 345696, HeapAlloc: 344784
CurrAlloc: 346200, HeapAlloc: 344784
CurrAlloc: 345288, HeapAlloc: 344784
CurrAlloc: 345792, HeapAlloc: 344784
CurrAlloc: 345224, HeapAlloc: 344784
CurrAlloc: 345696, HeapAlloc: 344784
CurrAlloc: 345192, HeapAlloc: 344784
CurrAlloc: 345696, HeapAlloc: 344784
What do you see instead?
CurrAlloc and VSS of the process grow without bounds
Which compiler are you using (5g, 6g, 8g, gccgo)?
Confirmed on 8g and 6g
Which operating system are you using?
Darwin and Linux
Which revision are you using? (hg identify)
All revision since http://code.google.com/p/go/source/detail?r=7c59a8fa31c3 are believed
affected.
Please provide any additional information below.
I believe the change which introduced this regression was
http://golang.org/cl/4291064. It is possible that the changes made in this CL
to improve channel allocation were not backed by a corresponding change in deallocation.
This is related to the issue https://golang.org/issue/1537 and
additional discussion at
http://groups.google.com/group/golang-nuts/browse_thread/thread/93b42ac1f888a615