[
https://issues.apache.org/jira/browse/CAMEL-26914?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18144465#comment-18144465
]
Claus Ibsen commented on CAMEL-26914:
-------------------------------------
What is left of this ticket after PR #29804
(https://github.com/apache/camel/pull/29804), which fixes B1 and B4, and B2 and
B3 for GitHub and gists:
* *B2 (rest)*: {{PipeProvider.java:112}} ({{camel bind}}) still fetches with
{{URL.openStream()}}, so with no connect or read timeout.
* *B3 (rest)*: {{HttpRequest.timeout()}} does not bound the response body, also
in {{ExampleHelper:553}}, {{VersionList:574}}, {{QuarkusHelper:348}},
{{ai/DependencyLookup:72}}, the launcher {{ManifestFetcher:218}} and
{{InstallScriptFetcher:102}}, and the streaming of {{LlmClient}} ({{camel ask}}
mid-answer, no idle timeout).
* *B4 (rest)*: {{install.sh}} runs {{curl}}/{{wget}} without a connect or max
time (the CLI now gives up after 5 minutes, but the installer has no limit of
its own).
* *B6*: {{camel infra run --background}} loses the output of the child
({{InfraRun.java:173-186}}, default pipes, the parent returns): with Docker
down the user only sees "Running in background".
* *B9*: {{camel debug}} reads stderr only after exit ({{Debug.java:459, 565,
614}}), so it can block on a full pipe, and {{doReadLog}} spins at 100% CPU
after EOF.
* *B10*: Vertex AI runs {{gcloud auth print-access-token}} on every LLM request
({{LlmClient.java:1704-1748}}): about 1 s per call, no timeout, and stderr is
merged into the token.
* *B11*: {{QuarkusHelper}} throws when the daily refresh fails offline, instead
of using the stale cache it has.
* *C9*: {{camel version list}} makes one sequential, uncached GET per unknown
version, before the {{--lts}}/{{--patch}} filters, and misses are never cached.
* *C10*: the GitHub example download ({{camel run --example}}, TUI) is
sequential and skips failed files silently.
_Claude Code on behalf of Claus Ibsen (davsclaus)_
> camel-jbang - outbound network calls: GitHub errors are swallowed, fetches
> without timeouts, and a stalled response body hangs past the timeout
> -----------------------------------------------------------------------------------------------------------------------------------------------
>
> Key: CAMEL-26914
> URL: https://issues.apache.org/jira/browse/CAMEL-26914
> Project: Camel
> Issue Type: Bug
> Components: camel-jbang
> Reporter: Claus Ibsen
> Assignee: Guillaume Nodet
> Priority: Major
> Fix For: 4.24.0
>
>
> Found by an analysis of the Camel CLI on the night of 2026-11-12, with the
> unpacked 4.24.0-SNAPSHOT camel-launcher on macOS. *(exp)* means reproduced by
> experiment, *(read)* means traced in the code. Paths are relative to
> {{dsl/camel-jbang/}}, and line numbers are from main at that time. Dependency
> resolution of camel run is covered separately by CAMEL-26901.
> h3. 1. GitHub and gist listings swallow non-200 responses (exp)
> {{GitHubHelper}} (camel-jbang-run {{common/GitHubHelper.java}} around lines
> 157-188) and {{GistHelper}} (around lines 88-121) ignore the status code.
> With a 404, or a 403/429 rate limit (60 requests an hour unauthenticated, and
> {{GITHUB_TOKEN}} is not supported), {{camel init <url>}} writes nothing and
> exits 0, and {{camel run <url>}} says "No routes found". They should report
> the status (and the rate limit), exit non-zero, and send {{GITHUB_TOKEN}}
> when it is set.
> h3. 2. Fetches with no timeouts (read)
> {{URL.openStream()}} with no connect or read timeout:
> * components/camel-resourceresolver-github {{GitHubResource}} and
> {{GistResource}}, used by camel run and camel init
> * {{PipeProvider}} of camel bind (around line 112)
> {{exists()}} swallows every error as "does not exist". The core
> {{HttpResource}} got timeouts in CAMEL-24756; these did not.
> h3. 3. The JDK HttpRequest.timeout() does not bound the response body (exp)
> A server that sends the headers and then stalls hangs
> {{send(BodyHandlers.ofString())}} past the timeout (still blocked after 18s
> with a 3s timeout). Affected:
> * {{GitHubHelper}}:158, {{GistHelper}}:89, {{ExampleHelper}}:553,
> {{VersionList}}:574, {{QuarkusHelper}}:348, ai {{DependencyLookup}}:72
> * launcher {{ManifestFetcher}}:218 and {{InstallScriptFetcher}}:102
> * {{LlmClient}} streaming: {{camel ask}} can hang mid-answer, there is no
> idle timeout
> Suggested fix: {{sendAsync}} plus {{get(timeout)}}, or a body handler with a
> deadline.
> h3. 4. camel self-update has no time limit and no output (read)
> {{SelfUpdateCommand}} (around line 369) uses {{readAllBytes()}} plus an
> untimed {{waitFor()}}, and install.sh runs {{curl}}/{{wget}} without a
> connect or max time.
> h3. 5. camel infra run --background loses the child's output (read)
> {{InfraRun}} (around lines 173-186) starts the child with the default pipes
> and returns. If Docker is down, the user only sees "Running in background".
> h3. 6. camel debug reads stderr only after exit (read)
> {{Debug.java}} (around lines 459, 565, 614) can block on a full stderr pipe,
> and {{doReadLog}} spins at 100% CPU after EOF.
> h3. 7. Vertex AI runs gcloud auth print-access-token on every LLM request
> (read)
> {{LlmClient}} (around lines 1704-1748): each call costs about 1s, has no
> timeout, and merges stderr into the token. The token should be cached until
> it expires.
> h3. 8. QuarkusHelper offline (read)
> When the daily refresh fails, it throws instead of using the stale cache it
> has.
> h3. 9. camel version list makes one sequential, uncached GET per unknown
> version (read)
> They run before the {{--lts}}/{{--patch}} filters, and misses are never
> cached.
> h3. 10. GitHub example downloads (read)
> The download of an example ({{camel run --example}}, and the TUI) is
> sequential and skips failed files silently. The TUI {{OsvClient}} also
> fetches vulnerability details one at a time.
> h3. Came out clean
> The Kubernetes plugin, the MCP server (no outbound HTTP), the HTTP and
> process handling of the TUI, and the core {{HttpResource}}. No security
> issues that need private triage were found.
> h3. Suggested order
> # 1 (silent failure)
> # 2, 3, 4 (hangs)
> # the rest
> _Claude Code on behalf of Claus Ibsen (davsclaus)_
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)