| Age | Commit message (Collapse) | Author |
|---|
|
Our cgit server has been shut down.
|
|
Use spawn with array to make SIGINT working effectively on Windows
|
|
which seems to prevent it from fetching notes when the path is not the
actual repository but a shallow-cloned repository.
|
|
|
|
|
|
from ruby/git.ruby-lang.org as of:
https://github.com/ruby/git.ruby-lang.org/commit/f3ed893e946ec66cac77af5859ac879c5983d3a3
|
|
|
|
from ruby/git.ruby-lang.org as of:
https://github.com/ruby/git.ruby-lang.org/commit/b0dfa734297cc9aea33f24a1e29f8853cc5761e9
|
|
Now envutil.rb is a part of test-unit-ruby-core gem, which still
supports old versions, 2.3 or later.
|
|
|
|
Empty pushref is no longer given 9b5eb828025cf0b7c023325e6e086422bcada355
|
|
Fixes 21c7131df818c1f7f571d4ccf9be150d2c9cc374 differently.
This makes it less likely that the printed command is deviated from
what's actually executed.
|
|
ZJIT: Improve zjit_bisect.rb
1. The command name in the help message is outdated.
2. When the command failed by itself, the message can be clearer.
|
|
|
|
|
|
|
|
Print a concatenated string instead of `print`ing multiple strings,
since `$\` is set to `"\n"` because of `-l` command line option.
|
|
|
|
No one uses `sync_default_gems_with` on other platforms; it is used by
the dedicated workflow and a few developers only.
|
|
|
|
When `-F` option is given to strip comments in bundled_gems file, `$;`
is set to the regexp.
On the other hand, old net/http.rb contained old CVS keyword expansion
that is `split` with the default separator, and non-nil `$;` causes a
warning.
```ruby
Revision = %q$Revision$.split[1]
```
|
|
|
|
|
|
instead of mutating RubyVM::Instructions and letting the order of
`require` impact its behavior.
Now that we have both RubyVM::TraceInstruction and
RubyVM::ZJITInstruction, it feels too much of a tight coupling to rely
on `require` to be ordered properly.
|
|
|
|
|
|
|
|
|
|
For example, `defined?(yield)` never calls a method, so it's leaf.
|
|
83bc6ced920c51ffc3783b03f2e1c7f89f5c2fd5 was the last use of it.
|
|
It was used to let MJIT override the leafness of the instruction when it
decides to remove check_ints for it. Now that MJIT is gone, nobody needs
to "override" the leafness using this.
|
|
MJIT was probably the last user of it. It's not even part of
build dependencies now.
|
|
|
|
|
|
|
|
```
Error: test_join(PathnameInstanceTest): NoMethodError: undefined method 'respond_to?' for an instance of RBS::UnitTest::Convertibles::ToStr
/Users/hsbt/Documents/github.com/ruby/ruby/gems/src/rbs/lib/rbs/unit_test/spy.rb:94:in 'Pathname#join'
/Users/hsbt/Documents/github.com/ruby/ruby/gems/src/rbs/lib/rbs/unit_test/spy.rb:94:in 'block (2 levels) in wrapped_object'
/Users/hsbt/Documents/github.com/ruby/ruby/gems/src/rbs/lib/rbs/unit_test/type_assertions.rb:152:in 'block in RBS::UnitTest::TypeAssertions#send_setup'
/Users/hsbt/Documents/github.com/ruby/ruby/gems/src/rbs/lib/rbs/unit_test/type_assertions.rb:150:in 'Kernel#catch'
/Users/hsbt/Documents/github.com/ruby/ruby/gems/src/rbs/lib/rbs/unit_test/type_assertions.rb:150:in 'RBS::UnitTest::TypeAssertions#send_setup'
/Users/hsbt/Documents/github.com/ruby/ruby/gems/src/rbs/lib/rbs/unit_test/type_assertions.rb:172:in 'RBS::UnitTest::TypeAssertions#assert_send_type'
test/stdlib/Pathname_test.rb:389:in 'PathnameInstanceTest#test_join'
386: Pathname('.'), :join, 'foo'
387: assert_send_type '(String, String) -> Pathname',
388: Pathname('.'), :join, 'foo', 'bar'
=> 389: assert_send_type '(ToStr) -> Pathname',
390: Pathname('.'), :join, ToStr.new('foo')
391: assert_send_type '(Pathname) -> Pathname',
392: Pathname('.'), :join, Pathname('foo')
```
|
|
|
|
|
|
Update included file list automatically.
|
|
|
|
|
|
|
|
|
|
The VPATH rule of NMake is different from others. Abandon using
them in the rules for the generated source, locate them in the top
source directory, as well as the generated library files of prism.
|
|
I spent a long time bisecting some largish program before realizing
that failure was in fact coming from the child timing out due to
zjit_bisect.rb not clearing the stdout and stderr pipe.
Fix the pipe deadlock by redirecting to /dev/null. This sacrifices
the debug output during boot, but for that we also get to remove
a lot of code and the dependency on Open3.
|
|
For most tests (except two), we don't need to change Encoding.default_internal
in assert_raise_with_message. We're trying to run the test suite across
multiple Ractors and modifying Encoding.default_internal can cause other
concurrently running tests to fail.
|
|
* Use unreleased version of rbs
* Skip all failing tests on windows
|
|
|
|
Also add a check in the bisect script that can assign blame to the HIR
optimizer.
|
|
|
