Setting warning-policy to "warn" still generates errors.

[stevenj]

When I run mdbook build with linkcheck enabled and warning-policy set to "warn" it still errors and returns a 101 error code? How do I configure it to just warn?

$ cat book.toml; mdbook build; echo $?
[book]
authors = ["Steven Johnson ([email protected])"]
language = "en"
multilingual = false
src = "src"
title = "Project Catalyst Standards"

[build]
build-dir = "book" # This is where github wants it so we can publish as gh-pages.

[preprocessor.regex]
patterns = "./src/assets/patterns/patterns.json"
before = [ # Make sure regex pre-processing happens before any other pre-processing.
    "kroki-preprocessor",
    "admonish",
    "open-on-gh"
]

[preprocessor.kroki-preprocessor]

[preprocessor.admonish]
command = "mdbook-admonish"
assets_version = "2.0.0" # do not edit: managed by `mdbook-admonish install`
on_failure = "bail"
before = [
    "kroki-preprocessor" # Ensure we can put diagrams inside admonish blocks.
]


[preprocessor.open-on-gh]
command = "mdbook-open-on-gh"
renderer = ["html"]

[output.html]
additional-css = [
    "./src/assets/css/mdbook-admonish.css",
    "./src/assets/css/open-in.css"
]

# Required: Your repository URL used in the link.
git-repository-url = "https://github.com/input-output-hk/catalyst-standards"

# Your git branch. Defaults to `main`
git-branch = "main"

# The text to use in the footer.
# The link text is marked by `[]`
open-on-text = "Found a bug? [Edit this page on GitHub.]"

[output.html.search]
enable = true            # enables the search feature
limit-results = 30       # maximum number of search results
teaser-word-count = 30   # number of words used for a search result teaser
use-boolean-and = true   # multiple search terms must all match
boost-title = 2          # ranking boost factor for matches in headers
boost-hierarchy = 1      # ranking boost factor for matches in page names
boost-paragraph = 1      # ranking boost factor for matches in text
expand = true            # partial words will match longer terms
heading-split-level = 3  # link results to heading levels
copy-js = true           # include Javascript code for search

[output.linkcheck]
# Should we check links on the internet? Enabling this option adds a
# non-negligible performance impact
follow-web-links = true

# Are we allowed to link to files outside of the book's root directory? This
# may help prevent linking to sensitive files (e.g. "../../../../etc/shadow")
traverse-parent-directories = false

# If necessary, you can exclude one or more links from being checked with a
# list of regular expressions. The regex will be applied to the link href (i.e.
# the `./index.html` in `[some page](./index.html)`) so it can be used to
# ignore both web and filesystem links.
#
# Hint: you can use TOML's raw strings (single quote) to avoid needing to
# escape things twice.
# exclude = [ 'google\.com' ]

# The User-Agent to use when sending web requests
user-agent = "mdbook-linkcheck-0.7.7"

# The number of seconds a cached result is valid for (12 hrs by default)
cache-timeout = 43200

# How should warnings be treated?
#
# - "warn" will emit warning messages
# - "error" treats all warnings as errors, failing the linkcheck
# - "ignore" will ignore warnings, suppressing diagnostic messages and allowing
#   the linkcheck to continuing
# We do NOT allow broken links in CI, but only warn in development.
# DO NOT EDIT THIS SETTING OR YOu WILL BREAK CI.
warning-policy = "warn"

# Extra HTTP headers that must be send to certain web sites
# in order to link check to succeed.
#
# This is a dictionary (map), with keys being regexes
# matching a set of web sites, and values being an array of
# the headers.
[output.linkcheck.http-headers]
# Any hyperlink that contains this regexp will be sent
# the "Accept: text/html" header
'crates\.io' = ["Accept: text/html"]

# mdbook-linkcheck will interpolate environment variables into your header via
# $IDENT.
#
# If this is not what you want you must escape the `$` symbol, like `\$TOKEN`.
# `\` itself can also be escaped via `\\`.
#
# Note: If interpolation fails, the header will be skipped and the failure will
# be logged. This can be useful if a particular header isn't always necessary,
# but may be helpful (e.g. when working with rate limiting).
#'website\.com' = ["Authorization: Basic $TOKEN"]2023-01-13 14:34:24 [INFO] (mdbook::book): Book building has started
2023-01-13 14:34:24 [INFO] (mdbook::book): Running the html backend
2023-01-13 14:34:24 [INFO] (mdbook::book): Running the linkcheck backend
2023-01-13 14:34:24 [INFO] (mdbook::renderer): Invoking the "linkcheck" renderer
error: File not found: ./doesnt_exist.html
  ┌─ chapter_1.md:3:1
  │
3 │ [broken link](./doesnt_exist.html)
  │ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File not found: ./doesnt_exist.html

Error: One or more incorrect links
2023-01-13 14:34:24 [ERROR] (mdbook::renderer): Renderer exited with non-zero return code.
2023-01-13 14:34:24 [ERROR] (mdbook::utils): Error: Rendering failed
2023-01-13 14:34:24 [ERROR] (mdbook::utils):    Caused By: The "linkcheck" renderer failed
101

I know its picking up the warning-policy option because if I make it an invalid value mdbook-linkcheck complains.

[jonaprieto]

+1 I have the same issue. It's quite bad in the scenario that a Github CI workflow deployment depends on
the right behavior of this option.

[paulcadman]

After a quick look at the code it looks like the add_invalid_link_diagnostics function does not respect the warning_policy, it always generates an error:

mdbook-linkcheck/src/validate.rs

Line 285 in dc22a02

let diag = Diagnostic::error()

Potential solution is to follow the pattern from add_incomplete_link_diagnostics instead:

mdbook-linkcheck/src/validate.rs

Line 270 in dc22a02

let diag = Diagnostic::new(severity)