Replace use of `STDIN`, `STDOUT` and `STDERR` constants by their
`$stdin`, `$stdout` and `$stderr` global variable equivalents.
This enables easier testing via standard means, such as `assert_output`
for minitest or `expect { print 'foo' }.to output.to_stdout` for RSpec
test suites.
a0a6fc1b76
* 🐛 Fixes [Bug #20039](https://bugs.ruby-lang.org/issues/20039)
When a Regexp is initialized with another Regexp, we simply copy the
properties from the original. However, the flags on the original were
not being copied correctly. This caused an issue when the original had
multibyte characters and was being compared with an ASCII string.
Without the forced encoding flag (`KCODE_FIXED`) transferred on to the
new Regexp, the comparison would fail. See the included test for an
example.
Co-authored-by: Nobuyoshi Nakada <nobu@ruby-lang.org>
Blocks should always look at their own local table first, even when
defined inside an ensure/rescue or something else that uses depth
offset. We can ignore the depth offset if we're doing local lookups
inside a block
This is based on the list in Gem::UpdateSuggestion and Bundler::Fetcher;
these have similar purposes (determining whether/what CI we're executing
in), and can benefit from being combined and updated (they're both
slightly out of date).
Noteable changes:
* We'll consider ourselves to be on a CI in more cases - if CI_NAME or
TASKCLUSTER_ROOT_URL are set specifically, since those represent two
cases that were either overlooked or are no longer covered by the
existing implementation. (Or possibly TaskCluster still does provide
RUN_ID, but failed to document it)
* We will notice/track a few additional services in ci_strings (cirrus,
dsari, taskcluster), stop tracking 'snap' (they went under in 2017),
and update buildbox to buildkite (they've been called that for 8
years, and the BUILDBOX envs have been deprecated for 3).
* We'll also sort/uniq/downcase the values (all of which only matter
because of the special case of CI_NAME).
60652b942f
Objects with the same shape must always have the same "embeddedness"
(either embedded or heap allocated) because YJIT assumes so. However,
using remove_instance_variable, it's possible that some objects are
embedded and some are heap allocated because it does not re-embed heap
allocated objects.
This commit changes remove_instance_variable to re-embed Object
instance variables when it becomes small enough.
Prior to this commit, we were not accounting for the case of a nil
parent in a CallXPathWriteNode, for example ::A ||= 1. This commit
checks if the parent exists, and if not, uses Object as the inferred
parent
The locals_body_index gives the index in the locals array where
the locals from the body start. This allows compilers to easily
index past the parameters in the locals array.
5d4627b890
In an array for `defined?` we need to check if there is a
`contains_splat` flag, if so bail early.
Ruby code:
```ruby
defined?([[*1..2], 3, *4..5])
```
Instructions:
```
"********* Ruby *************"
== disasm: #<ISeq:<compiled>@<compiled>:0 (0,0)-(0,29)>
0000 putobject "expression"
0002 leave
"********* PRISM *************"
== disasm: #<ISeq:<compiled>@<compiled>:0 (0,0)-(0,29)>
0000 putobject "expression"
0002 leave
```
It seems that when DNS connection is enabled, but the TCP connection is
disabled in a way, the `Net::HTTP.connect` raises `Net::OpenTimeout`.
And I want to skip the tests in this case.
042faf74e7/lib/net/http.rb (L1601-L1608)
```
1) Error:
TestBundledCA#test_accessing_new_index:
Net::OpenTimeout: execution expired
/builddir/build/BUILD/ruby-3.0.4/lib/net/http.rb:987:in `initialize'
/builddir/build/BUILD/ruby-3.0.4/lib/net/http.rb:987:in `open'
/builddir/build/BUILD/ruby-3.0.4/lib/net/http.rb:987:in `block in connect'
/builddir/build/BUILD/ruby-3.0.4/lib/timeout.rb:107:in `timeout'
/builddir/build/BUILD/ruby-3.0.4/lib/net/http.rb:985:in `connect'
/builddir/build/BUILD/ruby-3.0.4/lib/net/http.rb:970:in `do_start'
/builddir/build/BUILD/ruby-3.0.4/lib/net/http.rb:959:in `start'
/builddir/build/BUILD/ruby-3.0.4/lib/net/http.rb:1512:in `request'
/builddir/build/BUILD/ruby-3.0.4/lib/net/http.rb:1270:in `get'
/builddir/build/BUILD/ruby-3.0.4/test/rubygems/test_bundled_ca.rb:34:in `assert_https'
/builddir/build/BUILD/ruby-3.0.4/test/rubygems/test_bundled_ca.rb:58:in `test_accessing_new_index'
...
```
5defe0a2f6
It's possible to write the following and have it be valid Ruby:
```
defined?("foo"
)
```
But Prism wasn't taking the new line into account. This adds an
`accept1` for a `PM_TOKEN_NEWLINE` to account for this. I've also
updated the fixtures and snapshots to test this.
b87f8eedc6
Fix https://github.com/ruby/prism/pull/1468
Fix https://github.com/ruby/prism/pull/1575
To decide command-style method calls are allowed, this introduce a new
parameter `accepts_command_call` to `parse_expression` and some
functions.
Although one think this can be solved by operator precedence, it is
hard or impossible, because the precedence of command-style calls is skewed
(e.g. `! bar 1 ` is accepted, but `foo = ! bar 1` is rejected.)
One of the most complex examples is that:
(1) even though `foo = bar = baz 1` and `foo, bar = baz 1` is accepted,
(2) `foo, bar = baz = fuzz 1` is rejected.
To implement this behavior, this introduces a new binding power
`PM_BINDING_POWER_MULTI_ASSIGNMENT` and uses it for distinguish which single
assignments or multi assignments at their RHS.
d4dd49ca81
Previously only unknown word `intern` is allowed between a single-word
token before a C method. Now any single-word token, such as `inline`
which is used for `ArithmeticSequence` in enumerator.c, is allowed
instead.
3a214c1dd1
The change in #1055 might be a breaking change.
So, just simply wrap `token_stream` with `Array`
d8c19d7fa1
Co-authored-by: Jonathan Hefner <jonathan@hefner.pro>
Checking writability is prone to errors. For example:
$ mkdir -p foo/bar
$ chmod -w foo
$ touch foo/bar/baz # succeeds even if foo is not writable
6056138b6a
In #9101 I only accounted for an empty paren. This change implements the
`PM_PARENTHESES_NODE` for when it's `nil` and when it's an expression.
Code:
```ruby
defined?(("a"))
```
```
"********* Ruby *************"
== disasm: #<ISeq:<compiled>@<compiled>:0 (0,0)-(0,15)>
0000 putobject "expression"
0002 leave
"********* PRISM *************"
== disasm: #<ISeq:<compiled>@<compiled>:0 (0,0)-(0,15)>
0000 putobject "expression"
0002 leave
```
