Go 1.15: TestThriftSpanLogs errors

[eclipseo]

Go 1.15 rc 1 on Fedora Rawhide:

--- FAIL: TestThriftSpanLogs (0.00s)
    zipkin_thrift_span_test.go:264: 
        	Error Trace:	zipkin_thrift_span_test.go:264
        	Error:      	Not equal: 
        	            	expected: "{\"error\":\"drugs are baaad, m-k\"}"
        	            	actual  : "{\"error.object\":\"drugs are baaad, m-k\"}"
        	            	
        	            	Diff:
        	            	--- Expected
        	            	+++ Actual
        	            	@@ -1 +1 @@
        	            	-{"error":"drugs are baaad, m-k"}
        	            	+{"error.object":"drugs are baaad, m-k"}
        	Test:       	TestThriftSpanLogs
        	Messages:   	test-10
    zipkin_thrift_span_test.go:264: 
        	Error Trace:	zipkin_thrift_span_test.go:264
        	Error:      	Not equal: 
        	            	expected: "{\"error\":\"non-even keyValues len: 1\",\"function\":\"LogKV\"}"
        	            	actual  : "{\"error.object\":\"non-even keyValues len: 1\",\"function\":\"LogKV\"}"
        	            	
        	            	Diff:
        	            	--- Expected
        	            	+++ Actual
        	            	@@ -1 +1 @@
        	            	-{"error":"non-even keyValues len: 1","function":"LogKV"}
        	            	+{"error.object":"non-even keyValues len: 1","function":"LogKV"}
        	Test:       	TestThriftSpanLogs
        	Messages:   	test-14
FAIL
exit status 1
FAIL	github.com/uber/jaeger-client-go	3.711s

Each time the assert fail because error.object is returned instead of error.

[yurishkuro]

This isn't due to Go 1.15, but the newer version of OpenTracing. We should do a bump of OpenTracing in this repo too, and fix the tests.

[eclipseo]

Oh okay, sorry then, we always try to get the latest version in Fedora, sometimes it causes API breakage. Hopefully when we move to modules, we could have more fine tuning over dependencies version.

[yurishkuro]

@eclipseo it's fixed on master