* fix(startup-log): emit immediately after init and send to stderr

Startup logs now emit immediately after tracer initialization instead of
being delayed until the first payload is sent to the agent. They are also
sent to stderr (via console.error) instead of stdout, following Unix
conventions and aligning with other Datadog language tracers.

Changes:
- Added startupLog() call in proxy.js after initialization completes
- Changed output from info() to error() to use stderr
- Removed duplicate late call in writer.js that was firing on first payload
- Updated tests to expect console.error instead of console.info

Co-Authored-By: Claude Sonnet 4.5 <[email protected]>

* fix(tests): update data_streams_enabled tests to use console.error

After merging master, the new data_streams_enabled tests were still using
console.info instead of console.error. Updated all three tests in that
describe block to use console.error to match the stderr output change.

Co-Authored-By: Claude Sonnet 4.5 <[email protected]>

* fix(startup-log): separate initialization and agent diagnostic logs

Previously, the startup log and agent connection diagnostic were coupled
together in a single call. This meant that if the startup log was emitted
immediately (as intended), we would lose the agent connection diagnostic
that happens on the first write attempt.

Changes:
- Split startupLog() into two separate functions:
  - startupLog(): Logs tracer configuration immediately after init
  - logAgentError(): Logs agent connection diagnostic on first write error
- Each function maintains its own "already ran" flag to log only once
- Updated writer.js to call logAgentError() when agent responds with error
- Updated tests to verify both logs are emitted correctly and separately

This ensures users get both the immediate startup configuration log AND
the agent connection diagnostic if there's an issue reaching the agent.

Co-Authored-By: Claude Sonnet 4.5 <[email protected]>

* fix(startup-log): use warn for both startup log and agent diagnostic

Changed startup log to use warn() instead of error() for consistency.
Both the startup configuration log and agent diagnostic now use warn(),
which outputs to stderr via console.warn().

Changes:
- startup-log.js: Changed startupLog() to use warn() instead of error()
- Removed unused error import from log/writer
- Updated all tests to expect console.warn instead of console.error

This is more semantically appropriate since both logs are diagnostic
information rather than actual errors.

Co-Authored-By: Claude Sonnet 4.5 <[email protected]>

---------

Co-authored-by: Claude Sonnet 4.5 <[email protected]>

…err (#7470)"

This reverts commit 6bef168.

Changed startup logs to output to stderr (via console.warn) instead of
stdout (via console.info), following Unix conventions and aligning with
other Datadog language tracers.

This change does NOT modify when the logs are emitted - they continue to
be emitted on the first payload sent to the agent (same timing as before).

Changes:
- Separated startupLog() and logAgentError() into distinct functions
- Changed both functions to use warn() for stderr output
- Updated writer.js to call both functions separately
- Updated all tests to expect console.warn instead of console.info

Related to #7470 (which was reverted in #7478 due to regression from
immediate emission timing change). This PR includes only the stderr
output change without the timing modification.

Co-Authored-By: Claude Sonnet 4.5 <[email protected]>

Changed startup logs to output to stderr (via console.warn) instead of
stdout (via console.info), following Unix conventions and aligning with
other Datadog language tracers.

This is a minimal change that only swaps the output stream - no changes
to timing or function structure.

Changes:
- Changed startupLog() to use warn() instead of info() for stderr output
- Updated all tests to expect console.warn instead of console.info

Related to #7470 (which was reverted in #7478 due to regression from
immediate emission timing change). This PR includes only the stderr
output change without any timing modifications.

Co-Authored-By: Claude Sonnet 4.5 <[email protected]>

Changed startup logs to output to stderr (via console.warn) instead of
stdout (via console.info), following Unix conventions and aligning with
other Datadog language tracers.

This is a minimal change that only swaps the output stream - no changes
to timing or function structure.

Changes:
- Changed startupLog() to use warn() instead of info() for stderr output
- Updated all tests to expect console.warn instead of console.info

Related to #7470 (which was reverted in #7478 due to regression from
immediate emission timing change). This PR includes only the stderr
output change without any timing modifications.

Co-Authored-By: Claude Sonnet 4.5 <[email protected]>

* fix(startup-log): emit immediately after init and send to stderr

Startup logs now emit immediately after tracer initialization instead of
being delayed until the first payload is sent to the agent. They are also
sent to stderr (via console.error) instead of stdout, following Unix
conventions and aligning with other Datadog language tracers.

Changes:
- Added startupLog() call in proxy.js after initialization completes
- Changed output from info() to error() to use stderr
- Removed duplicate late call in writer.js that was firing on first payload
- Updated tests to expect console.error instead of console.info

Co-Authored-By: Claude Sonnet 4.5 <[email protected]>

* fix(tests): update data_streams_enabled tests to use console.error

After merging master, the new data_streams_enabled tests were still using
console.info instead of console.error. Updated all three tests in that
describe block to use console.error to match the stderr output change.

Co-Authored-By: Claude Sonnet 4.5 <[email protected]>

* fix(startup-log): separate initialization and agent diagnostic logs

Previously, the startup log and agent connection diagnostic were coupled
together in a single call. This meant that if the startup log was emitted
immediately (as intended), we would lose the agent connection diagnostic
that happens on the first write attempt.

Changes:
- Split startupLog() into two separate functions:
  - startupLog(): Logs tracer configuration immediately after init
  - logAgentError(): Logs agent connection diagnostic on first write error
- Each function maintains its own "already ran" flag to log only once
- Updated writer.js to call logAgentError() when agent responds with error
- Updated tests to verify both logs are emitted correctly and separately

This ensures users get both the immediate startup configuration log AND
the agent connection diagnostic if there's an issue reaching the agent.

Co-Authored-By: Claude Sonnet 4.5 <[email protected]>

* fix(startup-log): use warn for both startup log and agent diagnostic

Changed startup log to use warn() instead of error() for consistency.
Both the startup configuration log and agent diagnostic now use warn(),
which outputs to stderr via console.warn().

Changes:
- startup-log.js: Changed startupLog() to use warn() instead of error()
- Removed unused error import from log/writer
- Updated all tests to expect console.warn instead of console.error

This is more semantically appropriate since both logs are diagnostic
information rather than actual errors.

Co-Authored-By: Claude Sonnet 4.5 <[email protected]>

---------

Co-authored-by: Claude Sonnet 4.5 <[email protected]>

Changed startup logs to output to stderr (via console.warn) instead of
stdout (via console.info), following Unix conventions and aligning with
other Datadog language tracers.

This is a minimal change that only swaps the output stream - no changes
to timing or function structure.

Changes:
- Changed startupLog() to use warn() instead of info() for stderr output
- Updated all tests to expect console.warn instead of console.info

Related to #7470 (which was reverted in #7478 due to regression from
immediate emission timing change). This PR includes only the stderr
output change without any timing modifications.

Co-authored-by: Claude Sonnet 4.5 <[email protected]>

Changed startup logs to output to stderr (via console.warn) instead of
stdout (via console.info), following Unix conventions and aligning with
other Datadog language tracers.

This is a minimal change that only swaps the output stream - no changes
to timing or function structure.

Changes:
- Changed startupLog() to use warn() instead of info() for stderr output
- Updated all tests to expect console.warn instead of console.info

Related to #7470 (which was reverted in #7478 due to regression from
immediate emission timing change). This PR includes only the stderr
output change without any timing modifications.

Co-authored-by: Claude Sonnet 4.5 <[email protected]>

Changed startup logs to output to stderr (via console.warn) instead of
stdout (via console.info), following Unix conventions and aligning with
other Datadog language tracers.

This is a minimal change that only swaps the output stream - no changes
to timing or function structure.

Changes:
- Changed startupLog() to use warn() instead of info() for stderr output
- Updated all tests to expect console.warn instead of console.info

Related to #7470 (which was reverted in #7478 due to regression from
immediate emission timing change). This PR includes only the stderr
output change without any timing modifications.

Co-authored-by: Claude Sonnet 4.5 <[email protected]>

* fix(startup-log): emit immediately after init and send to stderr

Startup logs now emit immediately after tracer initialization instead of
being delayed until the first payload is sent to the agent. They are also
sent to stderr (via console.error) instead of stdout, following Unix
conventions and aligning with other Datadog language tracers.

Changes:
- Added startupLog() call in proxy.js after initialization completes
- Changed output from info() to error() to use stderr
- Removed duplicate late call in writer.js that was firing on first payload
- Updated tests to expect console.error instead of console.info

Co-Authored-By: Claude Sonnet 4.5 <[email protected]>

* fix(tests): update data_streams_enabled tests to use console.error

After merging master, the new data_streams_enabled tests were still using
console.info instead of console.error. Updated all three tests in that
describe block to use console.error to match the stderr output change.

Co-Authored-By: Claude Sonnet 4.5 <[email protected]>

* fix(startup-log): separate initialization and agent diagnostic logs

Previously, the startup log and agent connection diagnostic were coupled
together in a single call. This meant that if the startup log was emitted
immediately (as intended), we would lose the agent connection diagnostic
that happens on the first write attempt.

Changes:
- Split startupLog() into two separate functions:
  - startupLog(): Logs tracer configuration immediately after init
  - logAgentError(): Logs agent connection diagnostic on first write error
- Each function maintains its own "already ran" flag to log only once
- Updated writer.js to call logAgentError() when agent responds with error
- Updated tests to verify both logs are emitted correctly and separately

This ensures users get both the immediate startup configuration log AND
the agent connection diagnostic if there's an issue reaching the agent.

Co-Authored-By: Claude Sonnet 4.5 <[email protected]>

* fix(startup-log): use warn for both startup log and agent diagnostic

Changed startup log to use warn() instead of error() for consistency.
Both the startup configuration log and agent diagnostic now use warn(),
which outputs to stderr via console.warn().

Changes:
- startup-log.js: Changed startupLog() to use warn() instead of error()
- Removed unused error import from log/writer
- Updated all tests to expect console.warn instead of console.error

This is more semantically appropriate since both logs are diagnostic
information rather than actual errors.

Co-Authored-By: Claude Sonnet 4.5 <[email protected]>

---------

Co-authored-by: Claude Sonnet 4.5 <[email protected]>

…ent payload

Split startupLog into three independent functions so config is logged
immediately at init (via proxy.js) while integrations are logged later
on first agent payload (via writer.js). This gives users config info
faster without losing the integrations_loaded field that was regressed
in #7470.

Also fixes broken agent error logging introduced in #7212 where
`startupLog({ agentError: err })` didn't match the expected
`{ status, message }` shape.

Co-Authored-By: Claude Opus 4.6 <[email protected]>

#7643)

* fix(startup-log): emit config early at init, integrations on first agent payload

Split startupLog into three independent functions so config is logged
immediately at init (via proxy.js) while integrations are logged later
on first agent payload (via writer.js). This gives users config info
faster without losing the integrations_loaded field that was regressed
in #7470.

Also fixes broken agent error logging introduced in #7212 where
`startupLog({ agentError: err })` didn't match the expected
`{ status, message }` shape.

Co-Authored-By: Claude Opus 4.6 <[email protected]>

* ci: re-run CI

---------

Co-authored-by: Claude Opus 4.6 <[email protected]>

#7643)

* fix(startup-log): emit config early at init, integrations on first agent payload

Split startupLog into three independent functions so config is logged
immediately at init (via proxy.js) while integrations are logged later
on first agent payload (via writer.js). This gives users config info
faster without losing the integrations_loaded field that was regressed
in #7470.

Also fixes broken agent error logging introduced in #7212 where
`startupLog({ agentError: err })` didn't match the expected
`{ status, message }` shape.

Co-Authored-By: Claude Opus 4.6 <[email protected]>

* ci: re-run CI

---------

Co-authored-by: Claude Opus 4.6 <[email protected]>

#7643)

* fix(startup-log): emit config early at init, integrations on first agent payload

Split startupLog into three independent functions so config is logged
immediately at init (via proxy.js) while integrations are logged later
on first agent payload (via writer.js). This gives users config info
faster without losing the integrations_loaded field that was regressed
in #7470.

Also fixes broken agent error logging introduced in #7212 where
`startupLog({ agentError: err })` didn't match the expected
`{ status, message }` shape.

Co-Authored-By: Claude Opus 4.6 <[email protected]>

* ci: re-run CI

---------

Co-authored-by: Claude Opus 4.6 <[email protected]>