…l-adapter

…github.com:getsentry/sentry-javascript into lazarnikolov/js-2140-tanstack-start-tunnel-adapter

…l-adapter

…l-adapter

…l-adapter

Made-with: Cursor

Align sentryTanstackStart tests with TunnelRouteOptions.path and annotate
spy case with SentryTanstackStartOptions so options match the public API.

Made-with: Cursor

…l-adapter

Made-with: Cursor

…r quote styles

Made-with: Cursor

[Gitflow] Merge master into develop

After looking a bit at what the slowest browser-integration-tests are, I
found an easy-to-fix pattern in our tests - we had _a lot_ of tests
there that set some arbitrary `idleTimeout` to the
`browserIntegrationTest`. This massively made these tests slower as it
takes much longer to flush any pageload span. Many test runtimes reduced
from 10-20s to 1-5s due to this change. A few places remain where this
seems to be necessary, but there are way fewer now.

FYI 1000ms is the default, we also had some defining this for some
reason, I also removed those to be more consistent.

Use optional chaining on `.length` to collapse the redundant existence + length-guard ternary into a single expression.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Applies the same fix as in #19924

closes #18797

Co-authored-by: Julien Blatecky <julien.blatecky@hey.com>
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

This PR just isolates the mundane changes needed for a new SDK to keep
the next stacked PRs clean, it adds the Nitro SDK to the monorepo.

This PR is a base of a stack, the stacked PRs will be merged into it. I
thought this will be easier to review.

---

**This PR is part of a stack:**

- #20358
- #19224 👈
- #19225 
- #19304

---------

Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Covers the `{ path, allowedDsns }` shape of `tunnelRoute` end-to-end
alongside the existing dynamic/static/custom variants, exercising the
build-time serialization of `allowedDsns` into the generated virtual
route module. Also refactors the `tunnelRoute` resolution in
`vite.config.ts` from a nested ternary to a switch for readability.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

…l-adapter

The tunnel variants run independently through sentryTest.variants in
CI, so the chained `test:assert` was redundant and slow to run locally.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

…k-start-tunnel-adapter

Tunnel route helper + Dynamic tunnel route generator for TanStack Start React

There is no stable stream of Nitro 3 canary builds yet, so no point in
having this. I will add it back once we have a stable stream of builds,
likely after the full v3 release.

I probably had it at one point then didn't remove it properly before
merging #19224

Closes #20472

Add skill for writing tests. Created and evaluated with
https://github.com/anthropics/skills/blob/main/skills/skill-creator/SKILL.md

Used sources for "how to write good tests":
- https://vitest.dev/guide/learn/writing-tests-with-ai.html
- https://github.com/andredesousa/javascript-unit-testing-best-practices
- my knowledge from university - LOL

Basically #20527 but
for edge

closes #20368

…#20556)

We accidentally added a `node` dependency to an export from
`opentelemetry` package, leading to problems for users using this in a
browser environment.

This PR adds conditional exports to the opentelemetry package, where for
`browser` targets we have stubs for the node-only thing. This should
generally work the same as before, but stop failing builds in browser
envs.

I had to adjust browser integration tests for this a bit, as they did
some unnecessary aliasing which prevented webpack from using normal
conditional exports. We are simply doing less now and doing regular
dependency resolution which should work as expected (hopefully).

This updates the CODEOWNERS config to accomodate our new groups.

This PR adds instrumentation for LangGraph's `createReactAgent` API.

## createReactAgent wrapping
- Extracts agent name, LLM model, and tools from params
- Wraps compiled graph's `invoke()` with `invoke_agent` span
- Wraps tool `invoke()` with `execute_tool` spans (name, type,
description, arguments, result)
- Injects LangChain callback handler + `lc_agent_name` +
`__sentry_langgraph__` metadata at invoke level for chat span creation
and agent name propagation to all child spans
- Suppresses `StateGraph.compile` instrumentation inside
`createReactAgent` to avoid duplicate spans

## LangChain callback handler improvements
- Reads `gen_ai.agent.name` from `metadata.lc_agent_name` (convention
from newer LangGraph `createAgent`, adopted for our supported versions)
- Suppresses chain and tool callback spans inside agent context (based
on `metadata.__sentry_langgraph__` presence) to avoid duplicates with
our direct instrumentation
- Extracts tool definitions from `extraParams` in `handleChatModelStart`
and sets `gen_ai.request.available_tools` on chat spans
- Uses `runName` for tool name in `handleToolStart` (set by LangChain's
`StructuredTool.call()`) — fixes `unknown_tool` issue
- Adds `gen_ai.operation.name` to tool spans
- Extracts `.content` from ToolMessage objects in `handleToolEnd`
instead of serializing the full wrapper
- `addToolCallsAttributes` now prefers `message.tool_calls` (LangChain's
normalized format) over scanning `message.content` for Anthropic-style
`tool_use` items, fixing duplicate tool calls on Anthropic chat spans.
Falls back to `message.content` scanning for older LangChain versions.

## OTel module patching
- Patches `@langchain/langgraph/prebuilt` for `createReactAgent` (ESM +
CJS file patches for `dist/prebuilt/index.cjs`)

## Exports
- `instrumentCreateReactAgent` from core, browser, cloudflare

Closes: #19372

…SN` (#20528)

If this is set (which should generally be the case when using the
layer), we want to only allow this DSN to be forwarded.

If not set, it does not validate but warn that this is not validated.

---------

Co-authored-by: Andrei <168741329+andreiborza@users.noreply.github.com>

Saw this flake once or twice, just increasing timeout for cron tests to
allow for slow runner startup etc.

…ame (#20589)

Today, we got a bunch of flaky CI issues for e.g. failing cache or
playwright installation. While these are also kind of flakes, the issues
are not super actionable and it gets a bit noisy. I'd opt to not create
an issue if we cannot identify a concrete test that is flaking. We miss
out on some things being auto-reported here but that's also OK I'd say.

Closes #20319
Closes #20557
Closes #20493
Closes #20492
Closes #20469
Closes #20470

Fixes #20571

Closes #20268

…handlers (#20554)

Previously, Hono `.route()` was patched by checking for an `'ALL'`
handler (as those are used as middleware) to detect middleware in Hono
sub-apps. But this incorrectly treated `.all()` final handlers as
middleware. This broke error capture for `.all()` error routes (the
wrapped handler caught and marked the error, preventing responseHandler
from capturing it).

Now we use an arity heuristic (arity = number of params): middleware
accepts `(context, next)`, while route handlers accept only `(context)`.

Also added more middleware-related E2E test scenarios and route-patterns
tests. Basically adding test scenarios from the Node integration test to
the E2E test:
https://github.com/getsentry/sentry-javascript/blob/5d0d14531511fcd703438e072723ef98cd700ea3/dev-packages/node-integration-tests/suites/tracing/hono/test.ts

Part of Project:
#15260
Builds onto: #20449

Closes #20553

The clanker identified this as possible problem here:
> docker compose up --wait waits on the container healthcheck
(pg_isready inside the DB). On busy CI, localhost:5444 can still refuse
connections briefly afterward, so the scenario could hit Postgres before
it was reachable from the host and fail or behave inconsistently → flaky
CJS (and the same class of failure for the other scenarios).

Found some other places in browser-integration-tests that could
theoretically be flaky due to when we start listening for events vs.
when we navigate.