open-telemetry/opentelemetry-cpp

Summary BatchSpanProcessor and BatchLogRecordProcessor drain their queues during shutdown, but their timeout adjustment converts the default unlimited timeout into 1 microsecond before calling exporter shutdown. With asynchronous OTLP gRPC export enabled, exporter shutdown with 1 microsecond doesn't give enough time to wait for completion and cancels those requests. Steps to reproduce Configure trace and log providers with batch processors and OTLP gRPC exporters in YAML. Generate a small amount of trace and log data. Call ConfiguredSdk::UnInstall() to destroy the providers without explicitly calling ForceFlush() or setting a finite shutdown timeout. Expected Behavior Both a finite and unlimited shutdown timeouts should function as expected. The providers implicitly force flush where the data in the batch processor queues are drained and exported successfully. Observed Behavior Final exports are cancelled during shutdown due to the unlimited timeout being converted to a 1 us timeout.

The two build systems do not agree on which ext headers a consumer is allowed to include. CMake installs four, by name, and excludes the detail and server directories outright, in ext/CMakeLists.txt#L37-L51: FILES_MATCHING PATTERN "http_client.h" PATTERN "http_client_factory.h" PATTERN "http_client_factory_curl.h" PATTERN "url_parser.h" PATTERN "detail" EXCLUDE PATTERN "server" EXCLUDE) Bazel takes all ten, in ext/BUILD#L10: cc_library( name = "headers", hdrs = glob(["include/**/*.h"]), The six a Bazel consumer gets and a CMake consumer does not: ext/include/opentelemetry/ext/http/client/curl/http_client_curl.h ext/include/opentelemetry/ext/http/client/curl/http_operation_curl.h ext/include/opentelemetry/ext/http/client/curl/http_time_util.h ext/include/opentelemetry/ext/http/client/detail/default_factory.h ext/include/opentelemetry/ext/http/server/http_server.h ext/include/opentelemetry/ext/http/server/socket_tools.h Both halves compile. On b4257

Context While working on PR #4574, we found that sdk/include/opentelemetry/sdk/metrics/view/predicate_factory.h uses uint8_t without including <cstdint>. It compiled only because something else in the same translation unit happened to pull that header in transitively. This kind of bug is exactly what include-what-you-use (IWYU) checks are meant to catch, but it slipped through because this header has no matching .cc file of its own to anchor an IWYU check against. Proposal Look into extending the project's IWYU tooling/CI so that header-only files (headers with no corresponding .cc translation unit in the same directory) are also checked for missing or unnecessary includes, rather than relying on incidental coverage from whichever .cc file happens to include them first. Related #4574

Per open-telemetry/opentelemetry-specification#2983 and its fix, open-telemetry/opentelemetry-specification#5305, the metrics spec would require the periodic exporting MetricReader to collect and export whatever metrics were recorded since the last periodic tick before the underlying exporter is shut down. I'm not experienced in opentelemetry-cpp or its codebase, but the following AI-assisted analysis seems to confirm PeriodicExportingMetricReader::Shutdown doesn't currently do this: MetricReader::Shutdown sets the shutdown flag before calling OnShutDown. OnShutDown just wakes the worker thread, joins it, and shuts the exporter down — no extra collect. The worker's own loop exits on that wakeup without re-entering its collect-and-export step (CollectAndExportOnce), since the shutdown flag is checked after waking up, not before. OnForceFlush's wait condition returns immediately once IsShutdown() is true, without the worker having collected anything — so calling ForceFlush() manuall

Describe your environment Reproduced by source read + build on main at 11fa0db0 (also present in v1.28.0, the latest release -- not a regression). sdk/src/resource/resource.cc, Resource::Create. Steps to reproduce Resource Resource::Create(const ResourceAttributes &attributes, const std::string &schema_url) { ... if (resource.attributes_.find(semconv::service::kServiceName) == resource.attributes_.end()) { std::string default_service_name = "unknown_service"; auto it_process_executable_name = resource.attributes_.find(semconv::process::kProcessExecutableName); if (it_process_executable_name != resource.attributes_.end()) { default_service_name += ":" + nostd::get<std::string>(it_process_executable_name->second); } resource.attributes_[semconv::service::kServiceName] = default_service_name; } return resource; } it_process_executable_name->second is an AttributeValue (nostd::variant). nostd::get<std::string> throws if the variant doesn't cu