Bug when comparing size of len(msg.Pack) and msg.Len() #663
Comments
|
Here is a failing Test Case: func TestMsgCompressLengthLargeRecordsAllValues(t *testing.T) {
msg := new(Msg)
msg.Compress = true
msg.SetQuestion("redis.service.consul.", TypeSRV)
for i := 0; i < 900; i++ {
target := fmt.Sprintf("host-redis-%d-%d.test.acme.com.node.dc1.consul.", i/256, i%256)
msg.Answer = append(msg.Answer, &SRV{Hdr: RR_Header{Name: "redis.service.consul.", Class: 1, Rrtype: TypeSRV, Ttl: 0x3c}, Port: 0x4c57, Target: target})
msg.Extra = append(msg.Extra, &CNAME{Hdr: RR_Header{Name: target, Class: 1, Rrtype: TypeCNAME, Ttl: 0x3c}, Target: fmt.Sprintf("fx.168.%d.%d.", i/256, i%256)})
predicted := msg.Len()
buf, err := msg.Pack()
if err != nil {
t.Error(err)
}
if predicted != len(buf) {
t.Fatalf("predicted compressed length is wrong for %d records: predicted %s (len=%d) %d, actual %d", i, msg.Question[0].Name, len(msg.Answer), predicted, len(buf))
}
}
}It fails with the following: |
|
[ Quoting <notifications@github.com> in "[miekg/dns] Bug when comparing size..." ]
Following #635 I updated DNS lib to 1.0.5 into Consul but ran into another issue and I am still able to reproduce case where len(msg.Pack()) is not equal to msg.Len()
There could, maybe, be an one-off in Len(), but I don't have the time to deepdive
that atm.
|
|
@miekg I tried to find the issue... it seems it is related to the compression of data of Extra... I was not exactly able to find what is the issue, but this is definitly related to tests such as I'll try to find some time to dig a bit more |
|
[ Quoting <notifications@github.com> in "Re: [miekg/dns] Bug when comparing ..." ]
@miekg I tried to find the issue... it seems it is related to the compression of data of Extra... I was not exactly able to find what is the issue, but this is definitly related to tests such as `len < maxCompressionOffset`...
I'll try to find some time to dig a bit more
Thanks.
Yes, the `len < maxCompressionOffset` should maybe be `<=`, but I haven't given
this a lot of thought.
|
|
It took me a while to figure out the issue (a few hours), it was too simple :-) Since len was added before trying to add label to compression and since it is legal to have a start offset before 14 bits, we just had to more the increase of len at the end of loop :-) |
|
@miekg Ok, the patch I made seems to work fine now. I added lots of tests cases, including TestMsgCompressLengthLargeRecordsWithPaddingPermutation which test all permutations with padding around 14bits, so it should work well in all cases now |
Following #657 I updated DNS lib to 1.0.5 into Consul but ran into another issue and I am still able to reproduce case where len(msg.Pack()) is not equal to msg.Len()
pierresouchay/consul@f85db7e
It crashes with the following output:
I am gonna try to provide a test case
The text was updated successfully, but these errors were encountered: