The tests fail randomly on some platforms.
20220605T213004Z.fail.html.gz
20220605T210003Z.fail.html.gz
```
[15737/21701] TestThreadInstrumentation#test_thread_instrumentation_fork_safe/home/chkbuild/chkbuild/tmp/build/20220605T213004Z/ruby/tool/lib/test/unit/assertions.rb:109:in `assert': Expected 0 to be nonzero?. (Test::Unit::AssertionFailedError)
```
The failures seem to depend on context switches. I suspect `sleep 0.05`
is too short, so this change tries to extend the wait time.
Ref: https://bugs.ruby-lang.org/issues/18339
Design:
- This tries to minimize the overhead when no hook is registered.
It should only incur an extra unsynchronized boolean check.
- The hook list is protected with a read-write lock as to cause
contention when some hooks are registered.
- The hooks MUST be thread safe, and MUST NOT call into Ruby as they
are executed outside the GVL.
- It's simply a noop on Windows.
API:
```
rb_internal_thread_event_hook_t * rb_internal_thread_add_event_hook(rb_internal_thread_event_callback callback, rb_event_flag_t internal_event, void *user_data);
bool rb_internal_thread_remove_event_hook(rb_internal_thread_event_hook_t * hook);
```
You can subscribe to 3 events:
- READY: called right before attempting to acquire the GVL
- RESUMED: called right after successfully acquiring the GVL
- SUSPENDED: called right after releasing the GVL.
The hooks MUST be threadsafe, as they are executed outside of the GVL, they also MUST NOT call any Ruby API.
This issue was not detected because when all traces of old YAML parser
and emitter `Syck` were removed, this null-type.gemspec.rz marshalled
gemspec was updated to no longer load `YAML::Syck::PrivateType` but load
`Psych::PrivateType` instead.
However, realworld old marshalled gemspecs still use the
`YAML::PrivateType` constant, so this commit replaces the gemspec to be
the real pry-0.4.7 marshalled gemspec, so that it catches this issue.
51b330b8d2
This test is making sure that RubyGems is able to load old marshalled
gemspecs that include a field loading `YAML::PrivateType` instances in
the `rubyforge_project` field instead of `nil` due to a bug in old YAML
emitters.
At some point, the `rubyforge_project` field was removed and this test
was updated to just check another field. However, in the realworld other
fields do not have this issue and the marshalled gemspec we use for
testing still has this issue in the field reserved for the
`rubyforge_project` field. So I think updating the test to check other
field was not correct.
Instead, since the `rubyforge_project` field is now ignored, we no
longer need to worry about any conversion in this field. The only thing
we care about is that the marshal loading still works (which requires
that the constant is at least defined).
So this commit updates the test to make this more clear.
cddfacf6d4
`vm_trace_hook()` runs global hooks before running local hooks.
Previously, we read the local hook list before running the global hooks
which led to use-after-free when a global hook frees the local hook
list. A global hook can do this by disabling a local TracePoint, for
example.
Delay local hook list loading until after running the global hooks.
Issue discovered by Jeremy Evans in GH-5862.
[Bug #18730]
Try to mirror IO behavior, where chomp takes out the entire paragraph
separators between entries, but does not chomp a single line separator
at the end of the string.
Partially Fixes [Bug #18768]
a83ddbb7f0
Previously, this could result in an infinite loop. Always update
the e pointer in this case, setting w when chomping so the chomped
data is not included in the output.
Fixes [Bug #18769]
4bf64d5130
Fixes [Bug #18779]
Define the following methods as `rb_f_notimplement` on unsupported
platforms:
- GC.compact
- GC.auto_compact
- GC.auto_compact=
- GC.latest_compact_info
- GC.verify_compaction_references
This change allows users to call `GC.respond_to?(:compact)` to
properly test for compaction support. Previously, it was necessary to
invoke `GC.compact` or `GC.verify_compaction_references` and check if
those methods raised `NotImplementedError` to determine if compaction
was supported.
This follows the precedent set for other platform-specific
methods. For example, in `process.c` for methods such as
`Process.fork`, `Process.setpgid`, and `Process.getpriority`.
I am asking did_you_mean to use Exception#detailed_message to add
"Did you mean?" suggestion instead of overriding #message method.
https://github.com/ruby/did_you_mean/pull/177
Unfortunately, the change will affect Gem::UnknownCommandError, which
excepts did_you_mean to override #message method.
This PR absorbs the change of did_you_mean.
Gem::CommandManager now calls #detailed_message method to get a message
string with "Did you mean?" suggestion from an exception.
8f104228d3
When `HAVE_RB_ENC_INTERNED_STR` is enabled it is possible to
pass through a null pointer to `rb_enc_interned_str` resulting
in a segfault
Fixes#495b59368a8c2
This snuck in while addding tests for the `create_additions`
feature. Caught by JRuby when we added the `limit` option to the
Date/DateTime parsing methods, which causes this to be rejected as
an unknown keyword.
b1007dff66
This implements the getblockparam instruction.
There are two cases we need to handle depending on whether or not
VM_FRAME_FLAG_MODIFIED_BLOCK_PARAM is set in the environment flag.
When the modified flag is unset, we need to call rb_vm_bh_to_procval to
get a proc from our passed block, save the proc in the environment, and
set the modified flag.
In the case that the modified flag is set we are able to just use the
existing proc in the environment.
One quirk of this is that we need to call jit_prepare_routine_call early
and ensure we update PC and SP regardless of the branch taken, so that
we have a consistent SP offset at the start of the next instruction.
We considered using a chain guard to generate these two paths
separately, but decided against it because it's very common to see both
and the modified case is basically a subset of the instructions in the
unmodified case.
This includes tests for both getblockparam and getblockparamproxy which
was previously missing a test.
URI::Generic#{user,password} return the encoded values, which are
not that useful if you want to do authentication with them.
Automatic decoding by default would break backwards compatibility.
Optional automatic decoding via a keyword to URI.parse would
require threading the option through at least 3 other methods, and
would make semantics confusing (user= takes encoded or unencoded
password?) or require more work. Thus, adding this as a separate
method seemed the simplest approach.
Unfortunately, URI lacks a method for correct decoding. Unlike in
www form components, + in earlier parts of the URI such as the
userinfo section is treated verbatim and not as an encoded space.
Add URI.#{en,de}code_uri_component methods, which are almost the
same as URI.#{en,de}code_www_form_component, but without the
special SP => + handling.
Implements [Feature #9045]
16cfc4e92f
For string concat, see if compile-time encoding of strings matches.
If so, use simple buffer string concat at runtime. Otherwise, use
encoding-checking string concat.
Object#autoload implements a custom per-thread "mutex" for blocking
threads waiting on autoloading a feature. This causes problems when used
with the fiber scheduler. We swap the implementation to use a Ruby mutex
which is fiber aware.
