Socket test failing in Beats

[namrata-ibm]

While building Beats on a Big Endian system, one of the socket related test fails which uses Go sigar code sys/linux/inetdiag.go. Will it be possible to add Big Endian support in Go sigar ?

[namrata-ibm]

Any update on this?
@ruflin / @knz Could you please have a look?

[ruflin]

@andrewkroh is probably the better person here to give some feedback.

[andrewkroh]

Which test is failing? Can you provide the output?

The endianness is hardcoded in a few places IRRC. The endianness could be extracted to a variable and populated based on the detected endianness of the machine.

[namrata-ibm]

While executing go test -v github.com/elastic/beats/metricbeat/module/system/socket from metricbeat , the execution hangs at TestFetch:

=== RUN   TestHashSet
--- PASS: TestHashSet (0.00s)
=== RUN   TestListenerTable
--- PASS: TestListenerTable (0.00s)
=== RUN   TestData
--- SKIP: TestData (0.00s)
        data_generator.go:38: Skip data generation tests
=== RUN   TestFetch

If I modify the serialize and ParseInetDiagMsg functions from inetdiag.go file, to change the occurrences of binary.LittleEndian to binary.BigEndian, the test passes.

[namrata-ibm]

@andrewkroh, could you please have a look at the above?

[andrewkroh]

Oh, so it's a test in Metricbeat.

I think that package should have a variable like

var endianness binary.ByteOrder

and update all hard-coded places to use the variable.

Then have an init() function that detects the endian.

// NOT TESTED
func init() {
        var i int32 = 0x1
	v := (*[4]byte)(unsafe.Pointer(&i))
	if v[0] == 0 {
		endianness = binary.BigEndian
	} else {
		endianness = binary.LittleEndian
	}
}

WDYT?

[namrata-ibm]

@andrewkroh, ya this looks good.

[namrata-ibm]

Can this change be merged? @andrewkroh, the above patch works fine on little and big endian.

[Nayana-ibm]

@andrewkroh As you suggested in your comment, we have worked on adding big endian support for gosigar.
Raised PR #98 . Could you please review?

[Nayana-ibm]

As #100 is merged, Can you please update beats source code to have latest commit of gosigar? This will fix the issues with big endian.

[ruflin]

@andrewkroh Should we do a gosigar release?

[Nayana-ibm]

@andrewkroh @ruflin Can Beats use latest stable of gosigar ? This will fix the issues with big endian.

[andrewkroh]

It looks like v0.9.0 is being used by Beats on master and 6.3 branches. See vendor.json. v0.9.0 includes the endian changes.

[Nayana-ibm]

Thanks. Will wait for 6.3 tag