Module: Homebrew::Vulns::OsvExport Private

Defined in:
vulns/osv_export.rb

Overview

This module is part of a private API. This module may only be used in the Homebrew/brew repository. Third parties should avoid using this module if possible, as it may be removed or changed without warning.

Emits OSV-schema records for the Homebrew ecosystem describing CVEs that homebrew-core formulae resolve via shipped patches.

One record is written per (formula, vulnerability id) pair found in serialized_patches[].resolves. The record states that the formula was affected up to (but not including) the currently shipped version+revision; this is a "fixed at or before what we ship today" approximation, since the precise fix boundary requires homebrew-core git archaeology.

Record shape follows the OSV 1.7 schema and mirrors the Debian DSA layout (upstream listing the source CVE, affected[].ranges of type ECOSYSTEM, ecosystem_specific carrying the resolving patch detail).

See DevCmd::GenerateVulnsAdvisories for the entry point and https://github.com/Homebrew/advisory-database for the published feed.

Constant Summary collapse

SCHEMA_VERSION =

This constant is part of a private API. This constant may only be used in the Homebrew/brew repository. Third parties should avoid using this constant if possible, as it may be removed or changed without warning.

https://ossf.github.io/osv-schema/ — value of the emitted schema_version field, pinning the OSV schema release these records target. Homebrew and BREW were registered in that schema in ossf/osv-schema#576.

"1.7.3"
ECOSYSTEM =

This constant is part of a private API. This constant may only be used in the Homebrew/brew repository. Third parties should avoid using this constant if possible, as it may be removed or changed without warning. 

"Homebrew"
ID_PREFIX =

This constant is part of a private API. This constant may only be used in the Homebrew/brew repository. Third parties should avoid using this constant if possible, as it may be removed or changed without warning. 

"BREW"
PatchRef =

This constant is part of a private API. This constant may only be used in the Homebrew/brew repository. Third parties should avoid using this constant if possible, as it may be removed or changed without warning. 

T.type_alias { T::Hash[Symbol, T.any(String, T::Array[String])] }

Class Method Summary collapse

Class Method Details

.affected_entry(formula, vuln_id, patches, fixed) ⇒ Hash{Symbol => T.untyped}

This method is part of a private API. This method may only be used in the Homebrew/brew repository. Third parties should avoid using this method if possible, as it may be removed or changed without warning.

Parameters:

Returns:



444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
 
# File 'vulns/osv_export.rb', line 444

def self.affected_entry(formula, vuln_id, patches, fixed)
  {
    package:            {
      ecosystem: ECOSYSTEM,
      name:      formula.name,
      purl:      purl(formula.name),
    },
    ranges:             [
      {
        type:   "ECOSYSTEM",
        events: [{ introduced: "0" }, { fixed: }],
      },
    ],
    ecosystem_specific: {
      fix:     "patch",
      patches: patches_resolving(patches, vuln_id).filter_map { |p| patch_ref(p) },
    },
  }
end

.fetch_upstream(vuln_id) ⇒ Hash{String => T.untyped}, Symbol

This method is part of a private API. This method may only be used in the Homebrew/brew repository. Third parties should avoid using this method if possible, as it may be removed or changed without warning.

Returns :failed (not nil) on error so callers can distinguish a transient outage from a successful fetch that returned no enrichment.

Parameters:

Returns:



502
503
504
505
506
 
# File 'vulns/osv_export.rb', line 502

def self.fetch_upstream(vuln_id)
  OSV.vulnerability(vuln_id)
rescue OSV::Error
  :failed
end

.fixed_follows?(ranges, fixed) ⇒ Boolean

This method is part of a private API. This method may only be used in the Homebrew/brew repository. Third parties should avoid using this method if possible, as it may be removed or changed without warning.

True when fixed can close every reviewed open Homebrew ecosystem range. Terminal ranges are unchanged; invalid ranges remain eligible for the existing repair path.

Parameters:

  • ranges (T.untyped)
  • fixed (String)

Returns:

  • (Boolean) 


188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
 
# File 'vulns/osv_export.rb', line 188

def self.fixed_follows?(ranges, fixed)
  fixed_version = PkgVersion.parse(fixed)
  Array(ranges).all? do |range|
    next true unless range.is_a?(Hash)
    next true if (range["type"] || range[:type]) != "ECOSYSTEM"
    next true if range_state(range) != :open

    event = Array(range["events"] || range[:events]).rfind do |item|
      item.is_a?(Hash) && (item.key?("introduced") || item.key?(:introduced))
    end
    introduced = event&.[]("introduced") || event&.[](:introduced)
    introduced.is_a?(String) && fixed_version > PkgVersion.parse(introduced)
  rescue ArgumentError
    false
  end
rescue ArgumentError
  false
end

.merge_existing(path, record, close_open_ranges: false, initial_introduction: false) ⇒ Hash{Symbol => T.untyped}?

This method is part of a private API. This method may only be used in the Homebrew/brew repository. Third parties should avoid using this method if possible, as it may be removed or changed without warning.

If a record already exists at path, carry forward its published timestamp and affected[].ranges (so a terminal boundary does not drift to today's pkg_version), and skip the write entirely when nothing else has changed. close_open_ranges lets the matcher merge a newly discovered fixed or reintroduced event into an existing range while preserving its reviewed history. initial_introduction repairs unreviewed ranges without reopening valid terminal ranges. Records for annotations no longer in core are simply not visited, so they persist.

Parameters:

  • path (String)
  • record (Hash{Symbol => T.untyped})
  • close_open_ranges (Boolean) (defaults to: false)
  • initial_introduction (Boolean) (defaults to: false)

Returns:



112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
 
# File 'vulns/osv_export.rb', line 112

def self.merge_existing(path, record, close_open_ranges: false, initial_introduction: false)
  return record unless File.file?(path)

  existing = JSON.parse(File.read(path))
  # Records written before `published` was introduced only have
  # `modified`; use it as the migration value so `published` does not
  # jump forward to today on first rewrite.
  if (existing_published = existing["published"] || existing["modified"])
    record[:published] = existing_published
  end
  affected_records = Array(record[:affected])
  # Complete histories are not single boundary transitions. Only an
  # initial repair may copy their events into unreviewed ranges.
  if close_open_ranges && !initial_introduction &&
     affected_records.any? { |affected| multiple_intervals?(affected[:ranges]) }
    return
  end

  invalid_transition = affected_records.each_with_index.any? do |affected, index|
    existing_ranges = existing.dig("affected", index, "ranges")
    next false unless existing_ranges
    next false if initial_introduction && multiple_intervals?(affected[:ranges])

    close_open_ranges && (
      ((fixed = explicit_fixed(affected[:ranges])) && !fixed_follows?(existing_ranges, fixed)) ||
      (!initial_introduction && (introduced = explicit_reintroduction(affected[:ranges])) &&
        !reintroduction_follows?(existing_ranges, introduced))
    )
  end
  return if invalid_transition

  affected_records.each_with_index do |affected, index|
    existing_ranges = existing.dig("affected", index, "ranges")
    next unless existing_ranges

    affected[:ranges] = if close_open_ranges
      merge_range_transitions(existing_ranges, affected[:ranges], initial_introduction:)
    else
      existing_ranges
    end
  end

  # Compare as parsed structures so key ordering (which JSON does not
  # define but Ruby serialisation preserves) does not cause spurious
  # rewrites of a hand-formatted or differently-serialised existing file.
  return if JSON.parse(JSON.generate(record)).except("modified") == existing.except("modified")

  record
rescue JSON::ParserError
  record
end

.patch_ref(patch) ⇒ PatchRef?

This method is part of a private API. This method may only be used in the Homebrew/brew repository. Third parties should avoid using this method if possible, as it may be removed or changed without warning.

Parameters:

Returns:



490
491
492
493
494
495
496
497
 
# File 'vulns/osv_export.rb', line 490

def self.patch_ref(patch)
  ref = T.let({}, PatchRef)
  ref[:type] = patch["type"] if patch["type"]
  ref[:url] = patch["url"] if patch["url"]
  ref[:file] = patch["file"] if patch["file"]
  ref[:apply] = patch["apply"] if patch["apply"]
  ref.presence
end

.patches_resolving(serialized_patches, vuln_id) ⇒ Array<Hash{String => T.untyped}>

This method is part of a private API. This method may only be used in the Homebrew/brew repository. Third parties should avoid using this method if possible, as it may be removed or changed without warning.

Parameters:

Returns:



480
481
482
483
484
485
 
# File 'vulns/osv_export.rb', line 480

def self.patches_resolving(serialized_patches, vuln_id)
  target = vuln_id.upcase
  serialized_patches.select do |p|
    Array(p["resolves"]).any? { |r| r.is_a?(Hash) && r["type"] == "security" && r["id"].to_s.upcase == target }
  end
end

.purl(name) ⇒ String

This method is part of a private API. This method may only be used in the Homebrew/brew repository. Third parties should avoid using this method if possible, as it may be removed or changed without warning.

Parameters:

Returns:



472
473
474
 
# File 'vulns/osv_export.rb', line 472

def self.purl(name)
  "pkg:brew/#{name.gsub(/[@+]/, PURL_NAME_ENCODE)}"
end

.ranges_open?(ranges) ⇒ Boolean

This method is part of a private API. This method may only be used in the Homebrew/brew repository. Third parties should avoid using this method if possible, as it may be removed or changed without warning.

Parameters:

  • ranges (T.untyped)

Returns:

  • (Boolean) 


176
177
178
179
180
181
182
 
# File 'vulns/osv_export.rb', line 176

def self.ranges_open?(ranges)
  return false unless ranges.is_a?(Array)
  return false if ranges.empty?

  states = ranges.map { |range| range_state(range) }
  states.exclude?(:invalid) && states.include?(:open)
end

.ranges_terminal?(ranges) ⇒ Boolean

This method is part of a private API. This method may only be used in the Homebrew/brew repository. Third parties should avoid using this method if possible, as it may be removed or changed without warning.

True only when every range has a terminal fixed, last_affected, or limit event. Invalid/empty ranges deliberately return false so the selective matcher repairs them instead of treating them as reviewed.

Parameters:

  • ranges (T.untyped)

Returns:

  • (Boolean) 


168
169
170
171
172
173
 
# File 'vulns/osv_export.rb', line 168

def self.ranges_terminal?(ranges)
  return false unless ranges.is_a?(Array)
  return false if ranges.empty?

  !!ranges.all? { |range| range_state(range) == :terminal }
end

.record_for(formula, vuln_id, patches: formula.serialized_patches, fixed: formula.pkg_version.to_s, upstream: nil, now: Time.now.utc) ⇒ Hash{Symbol => T.untyped}

This method is part of a private API. This method may only be used in the Homebrew/brew repository. Third parties should avoid using this method if possible, as it may be removed or changed without warning.

Parameters:

  • formula (Formula)
  • vuln_id (String)
  • patches (Array<Hash{String => T.untyped}>) (defaults to: formula.serialized_patches)
  • fixed (String) (defaults to: formula.pkg_version.to_s)
  • upstream (Hash{String => T.untyped}, nil) (defaults to: nil)
  • now (Time) (defaults to: Time.now.utc)

Returns:



403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
 
# File 'vulns/osv_export.rb', line 403

def self.record_for(formula, vuln_id, patches: formula.serialized_patches,
                    fixed: formula.pkg_version.to_s, upstream: nil, now: Time.now.utc)
  timestamp = now.strftime("%Y-%m-%dT%H:%M:%SZ")
  record = T.let({
    schema_version:    SCHEMA_VERSION,
    id:                record_id(formula, vuln_id),
    published:         timestamp,
    modified:          timestamp,
    upstream:          [vuln_id],
    affected:          [affected_entry(formula, vuln_id, patches, fixed)],
    database_specific: { source: "generated" },
  }, T::Hash[Symbol, T.untyped])

  if upstream
    record[:summary] = upstream["summary"] if upstream["summary"]
    record[:details] = upstream["details"] if upstream["details"]
    record[:severity] = upstream["severity"] if upstream["severity"]
    record[:upstream] = ([vuln_id] + Array(upstream["aliases"])).uniq
    if (refs = upstream["references"])
      # OSV.dev merges NVD and cve.org reference lists without normalising
      # percent-encoding, so the same URL can appear twice (e.g. `%40` vs
      # `@`). Collapse those while keeping the same URL under distinct
      # `type` values, which the schema allows and which carries meaning.
      record[:references] = refs.uniq do |r|
        [r["type"], URI::RFC2396_PARSER.unescape(r["url"].to_s)]
      end
    end
  end

  record
end

.record_id(formula, vuln_id) ⇒ String

This method is part of a private API. This method may only be used in the Homebrew/brew repository. Third parties should avoid using this method if possible, as it may be removed or changed without warning.

Parameters:

Returns:



436
437
438
 
# File 'vulns/osv_export.rb', line 436

def self.record_id(formula, vuln_id)
  "#{ID_PREFIX}-#{formula.name}-#{vuln_id}"
end

.reintroduction_follows?(ranges, introduced) ⇒ Boolean

This method is part of a private API. This method may only be used in the Homebrew/brew repository. Third parties should avoid using this method if possible, as it may be removed or changed without warning.

True when introduced is compatible with every reviewed Homebrew ecosystem range. Open ranges are already affected and remain unchanged; terminal ranges must end before the new boundary. Other OSV range types (notably GIT commit ranges) are ignored.

Parameters:

  • ranges (T.untyped)
  • introduced (String)

Returns:

  • (Boolean) 


212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
 
# File 'vulns/osv_export.rb', line 212

def self.reintroduction_follows?(ranges, introduced)
  ecosystem_ranges = Array(ranges).select do |range|
    range.is_a?(Hash) && (range["type"] || range[:type]) == "ECOSYSTEM"
  end
  return false if ecosystem_ranges.empty?

  introduced_version = PkgVersion.parse(introduced)
  ecosystem_ranges.all? do |range|
    state = range_state(range)
    next true if state == :open
    next false if state != :terminal

    terminal = terminal_event(range)
    next false unless terminal

    terminal_key = TERMINAL_EVENT_KEYS.find do |key|
      terminal.key?(key) || terminal.key?(key.to_sym)
    end
    next false unless terminal_key

    terminal_version = terminal[terminal_key] || terminal[terminal_key.to_sym]
    next false unless terminal_version.is_a?(String)

    terminal_version = PkgVersion.parse(terminal_version)
    if terminal_key == "limit"
      introduced_version >= terminal_version
    else
      introduced_version > terminal_version
    end
  rescue ArgumentError
    false
  end
rescue ArgumentError
  false
end

.run(annotated, dir, first_fixed: nil, now: Time.now.utc) ⇒ Array<String>

This method is part of a private API. This method may only be used in the Homebrew/brew repository. Third parties should avoid using this method if possible, as it may be removed or changed without warning.

annotated is a list of [formula, serialized_patches] pairs. The patches are passed in rather than read from the formula so callers can supply the union across OS/architecture variations (a patch inside an on_linux/on_intel block only appears in Formula#serialized_patches under the matching SimulateSystem).

first_fixed, when given, is called (formula, vuln_id) -> String? for records with no existing file to derive an accurate fixed boundary (e.g. via FormulaVersions git history). It may return :history_unavailable to skip a new record when no boundary can be verified; existing records preserve their on-disk ranges regardless.

Parameters:

Returns:



60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
 
# File 'vulns/osv_export.rb', line 60

def self.run(annotated, dir, first_fixed: nil, now: Time.now.utc)
  FileUtils.mkdir_p(dir)
  written = []
  upstream_cache = T.let({}, T::Hash[String, T.any(T::Hash[String, T.untyped], Symbol)])

  annotated.each do |formula, patches|
    Scanner.resolved_ids(patches).each do |vuln_id|
      upstream = upstream_cache.fetch(vuln_id) { upstream_cache[vuln_id] = fetch_upstream(vuln_id) }
      path = File.join(dir, "#{record_id(formula, vuln_id)}.json")
      existing = File.file?(path)
      # A transient OSV outage would otherwise strip summary/severity/etc.
      # from an existing enriched record; leave it untouched instead.
      next if upstream == :failed && existing

      fixed_result = T.let(nil, T.nilable(T.any(String, Symbol)))
      fixed_result = first_fixed.call(formula, vuln_id) if first_fixed && !existing
      fixed = case fixed_result
      when String
        fixed_result
      when nil
        formula.pkg_version.to_s
      when :history_unavailable
        next
      else
        raise TypeError, "unexpected first-fixed result: #{fixed_result.inspect}"
      end
      record = record_for(formula, vuln_id, patches:, fixed:,
                          upstream: upstream.is_a?(Hash) ? upstream : nil, now:)
      merged = merge_existing(path, record)
      next if merged.nil?

      File.write(path, "#{JSON.pretty_generate(merged)}\n")
      written << path
    end
  end

  written
end