[TEST] Test failure in OtlpHttpExporterTestPeer.ExportJsonIntegrationTestSync

[marcalff]

[ RUN      ] OtlpHttpExporterTestPeer.ExportJsonIntegrationTestSync
exporters/otlp/test/otlp_http_exporter_test.cc:173: Failure
Mock function called more times than expected - returning directly.
    Function call: SendRequest((ptr = 0x7f65540106a0, value = 8-byte object <18-8A E1-5E 65-7F 00-00>))
         Expected: to be called once
           Actual: called twice - over-saturated and active

-- Test timed out at 2025-10-27 17:08:13 UTC --

The test exports 2 spans.
I suspect the batch record processor triggered an export between the two spans, leading to 2 exports.

Possible fix: replace WillOnce() with Times Between (1, 2), to account for 1 or 2 exports.

[ThomsonTan]

Curious on why it is missed in CI.

[marcalff]

This one was found in CI, in a PR #3715.

See https://github.com/open-telemetry/opentelemetry-cpp/actions/runs/18848842669/job/53779993301

Due to rebase, when CI is executed again and passes, these get less visibility.

[marcalff]

Seen again today on #3758

[ RUN      ] OtlpHttpExporterTestPeer.ExportJsonIntegrationTestAsync
exporters/otlp/test/otlp_http_exporter_test.cc:265: Failure
Mock function called more times than expected - returning directly.
    Function call: SendRequest((ptr = 0x7fad9801b550, value = 8-byte object <18-1A 4E-A3 AD-7F 00-00>))
         Expected: to be called once
           Actual: called twice - over-saturated and active

[  FAILED  ] OtlpHttpExporterTestPeer.ExportJsonIntegrationTestAsync (104 ms)

[owent]

Sometimes jitter in CPU time-slice allocation causes part of the data export to be split into two runs; I’ll change it later to support exporting the data in two parts.

[marcalff]

See related #3760

[owent]

See related #3760

Thanks. For OTLP, it's a bit more complex: we need to verify the data records even when the export is split into multiple batches.

Sorry, I've been very busy lately; I'll complete this in a few days.