[pull] master from ruby:master#1071
Merged
Merged
Conversation
Fold arithmetic identity operations such as `x + 0`, `x - 0`, `x * 1`, `x / 1` in `fold_constants`.
When VM state is corrupted enough, we can call abort() from the SIGABRT
handler. Previously, we would spam until the stack is full:
ABRT received in SEGV handler
SEGV received in ABRT handler
ABRT received in SEGV handler
ABRT received in ABRT handler
ABRT received in ABRT handler
ABRT received in ABRT handler
[...]
We've seen this on CI:
https://github.com/ruby/ruby/actions/runs/26591192708/job/78350130653
To test this situation locally, temporarily patch in a call to abort()
in rb_bug_for_fatal_signal() then use `Process.kill(:ABRT, Process.pid)`.
Fix by restoring the default signal handler before aborting.
The completeness check skips the download once every expected file exists, but File.write truncates before writing, so an install interrupted partway leaves a half-written file in place. A later run would see it, treat the vendor tree as complete, and load a truncated source. Download each file to a sibling .tmp path and rename it into place. The rename stays within the vendor tree, so it is atomic and a target file is only ever observed fully written. ruby/rubygems@a2f265f579 Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
When captured subprocess output contained a byte that isn't valid UTF-8, normalize tagged the buffer as UTF-8 and then ran gsub over it, which scans the whole string and raised Encoding::CompatibilityError: invalid byte sequence in UTF-8. The error surfaced from inside the test harness rather than on an assertion, so RSpec reported "Unable to find matching line from backtrace" and the real output was lost. Scrub the string after forcing the encoding so normalize never raises on malformed bytes, and dup it first so forcing the encoding no longer mutates the stored capture buffer. ruby/rubygems@e236edcb6d Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…ions Installing a git source clones the local cache repo into bundler/gems with git's hardlink optimization. Git 2.54 kicks off commit-graph writes as part of automatic maintenance after a fetch or clone, and when that background write touches the cache while the hardlinking clone is enumerating it, the clone aborts with "hardlink different from source" pointing at a tmp_graph_* file. It shows up as a flaky failure on CI. Pass gc.auto=0 and maintenance.auto=false to every git invocation so no background maintenance is ever spawned for these throwaway repos, removing the race. The options are injected only at the command-execution layer, so redacted command strings in error messages are unaffected. ruby/rubygems@7bc41a21fa Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
test_with_webauthn_enabled_failure checked the Authorization header on @stub_fetcher.last_request, but the webauthn flow runs a real polling thread that issues its own request concurrently. On Rubies without a GIL the poll request can land last, so last_request was not the webauthn_verification request and the assertion saw a nil header. Look up the webauthn_verification request explicitly and assert on its Authorization header, which the main thread always records. This is the invariant the test means to check and it is no longer order dependent, so the TruffleRuby pend that papered over the same race can go too. ruby/rubygems@98efee308a Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
io_reader trusted IO#read to honour its size argument and copied RSTRING_LEN(string) bytes into libyaml's fixed-capacity buffer. An IO-like object whose #read over-returns wrote past the buffer end, a heap out-of-bounds write reachable from Psych.load/safe_load/parse. Clamp the copied length to the requested size. ruby/psych@99ecd94560 Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Clamping the over-read at exactly the requested size could split a multibyte character, since the string an IO returns may carry a non-binary encoding. Round the cut down to the last character boundary at or before the size so the bytes handed to libyaml are always whole characters. ruby/psych@6201ae17c1 Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
End procs are executed in reverse order of registration. Since the crypt provider is initialized very early on via `Init_RandomSeedCore`, no other end procs are executed after the crypt provider is released. Finalizers are called after end procs, but their order is unspecified, so there is still no guarantee that this crypt provider will be available for use by other end procs.
…st colon The upcoming compact index v2 format introduced by rubygems.org appends a `created_at:ISO8601` field to each info line, and the timestamp's internal colons were being treated as additional split delimiters. Splitting only on the first colon keeps existing keys like `ruby:>= 2.7.0` intact while letting multi-colon values pass through untouched. rubygems/rubygems.org#6504 ruby/rubygems@227b3d6f94 Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
The cooldown feature needs each gem version's publish timestamp on the client side. Compact index v2 exposes it as a `created_at:ISO8601` entry in the info-line metadata; expose a Time-typed `created_at` attribute on the spec so the resolver can consult it later. Parsing is defensive against older rubygems whose APISet GemParser splits ISO8601 timestamps on every colon, accepting both Array and flat String shapes and silently dropping malformed values. The time stdlib is required lazily inside the case branch so loading the file does not activate the `time` default gem during Bundler.setup. ruby/rubygems@cd61070cfc Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds `cooldown` to NUMBER_KEYS so that `BUNDLE_COOLDOWN` and `bundle config set cooldown` are parsed as integer days. Reading the value is enough for now; later commits plumb it into the resolver and the Gemfile DSL. ruby/rubygems#9113 ruby/rubygems@478c3ff09f Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Lets `source "https://rubygems.org", cooldown: 7` attach a per-remote value to the global Rubygems source, which the new `Remote#effective_cooldown` reads with CLI > config > Gemfile per-source precedence. The cooldown is stored on Source::Rubygems keyed by URI so that several top-level `source` lines can carry independent values onto the same global source. Non-negative Integer values are required at parse time; everything else (strings, floats, arrays, negative numbers) raises InvalidOption so a typo can't silently disable the filter. ruby/rubygems@5bd253f11f Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds the CLI flag and routes it through `set_command_option_if_given` so the value lands in Bundler.settings' temporary store. This keeps the CLI > config > Gemfile precedence implicit in the existing settings layering. The resolver does not yet consult the value. ruby/rubygems#9113 ruby/rubygems@b3a8b2e082 Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
…in errors Adds a per-version filter to `Resolver#filter_specs` that drops specs whose `created_at` falls within the effective cooldown window. The filter only runs over `@all_specs`, so lockfile-pinned versions in `@base` survive even if a newer release would now be excluded. Specs without `created_at` are kept so historical versions and indexes that do not yet expose timestamps remain usable. A shared `cooldown_now` memoization ensures every comparison within one resolve sees the same timestamp, stabilizing tests near a threshold boundary. When the resolver fails because every matching version is in cooldown, both `raise_not_found!` and `no_versions_incompatibility_for` surface a hint suggesting `--cooldown 0` to bypass; a small `cooldown_hint` helper keeps the wording in sync between the two error paths and is locked down with unit specs. ruby/rubygems@82dcf047ff Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
bundle outdated still surfaces newer-but-cooldown'd versions instead of hiding them, so the user knows an upgrade is pending rather than silently missing. The prose form gains ", in cooldown for Nd more days" and the table form appends "(cooldown Nd)" to the Latest column. ruby/rubygems@66b5ab78e0 Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds the `--cooldown=NUMBER` option to the install, update, add, and outdated man pages and describes the `cooldown` / `BUNDLE_COOLDOWN` setting in bundle-config, regenerating the rendered .1 files via `rake man:build`. ruby/rubygems@6f4b30491d Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
…d-to-end A `CompactIndexCooldownAPI` subclass overrides `build_gem_version` to emit `CompactIndex::GemVersionV2` with `created_at` sourced from `spec.date`, letting tests carry a deterministic timestamp per built gem. `spec/install/cooldown_spec.rb` exercises the source DSL keyword, `--cooldown` flag, `BUNDLE_COOLDOWN`, `bundle config set cooldown`, the rolling-delay filter, the lockfile bypass, and the CLI vs Gemfile precedence against the new artifice. ruby/rubygems@1f1309a809 Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
The Gemfile DSL already rejects `cooldown: -7`, but `--cooldown -7` on install/update/add/outdated slipped through Thor and ended up disabling the filter silently via the `days <= 0` short-circuit in the resolver. Add a shared CLI::Common.validate_cooldown! guard so the CLI surface fails loud with the same message as the DSL. ruby/rubygems@45e54d5b4c Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds three E2E checks against the v2 cooldown artifice: the negative CLI value is rejected at parse time, `bundle outdated --cooldown` tags the latest-but-cooled version in both table and parseable output, and `bundle update --cooldown 99999` surfaces the cooldown hint when every candidate is filtered. These were the remaining coverage gaps called out in the adversarial review. ruby/rubygems@3df6b300f7 Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
A full ISO-8601 timestamp with seconds (the format compact index v2 sends, e.g. 2026-05-30T08:52:10Z) is recognised by Time.new directly, so we can drop the require "time" and the indirection through Time.iso8601 while keeping the same ArgumentError fallback for malformed input. ruby/rubygems@d14e3ce964 Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
10 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
See Commits and Changes for more details.
Created by
pull[bot] (v2.0.0-alpha.4)
Can you help keep this open source service alive? 💖 Please sponsor : )