This PR fixes the following error for `Prism::Translation::Parser::Lexer` on the main branch:
```console
$ cat example.rb
'a' # aあ
"
#{x}
"
$ bundle exec rubocop
Parser::Source::Range: end_pos must not be less than begin_pos
/Users/koic/.rbenv/versions/3.0.4/lib/ruby/gems/3.0.0/gems/parser-3.3.0.5/lib/parser/source/range.rb:39:in `initialize'
/Users/koic/src/github.com/ruby/prism/lib/prism/translation/parser/lexer.rb:299:in `new'
/Users/koic/src/github.com/ruby/prism/lib/prism/translation/parser/lexer.rb:299:in `block in to_a'
/Users/koic/src/github.com/ruby/prism/lib/prism/translation/parser/lexer.rb:297:in `map'
/Users/koic/src/github.com/ruby/prism/lib/prism/translation/parser/lexer.rb:297:in `to_a'
/Users/koic/src/github.com/ruby/prism/lib/prism/translation/parser.rb:263:in `build_tokens'
/Users/koic/src/github.com/ruby/prism/lib/prism/translation/parser.rb:92:in `tokenize'
```
This change was made in https://github.com/ruby/prism/pull/2557, and it seems there was
an inconsistency in Range due to forgetting to apply `offset_cache` to `start_offset`.
Previously, using ToMarkdown on a label-list would generate output that
could not be reparsed by the RDoc::Markdown parser:
```
md = <<~MD
apple
: a red fruit
banana
: a yellow fruit
MD
doc = RDoc::Markdown.parse(md)
doc # => [doc: [list: NOTE [item: ["apple"]; [para: "a red fruit"]], [item: ["banana"]; [para: "a yellow fruit"]]]]
new_md = doc.accept(RDoc::Markup::ToMarkdown.new)
new_md # => "apple\n: a red fruit\nbanana\n: a yellow fruit\n\n"
new_doc = RDoc::Markdown.parse(new_md)
new_doc # => [doc: [list: NOTE [item: ["apple"]; [para: "a red fruit\nbanana\n: a yellow fruit"]]]]
```
The issue is that the [PHP Markdown Extra spec][1] requires a newline
after each definition list item, but ToMarkdown was not putting newlines
between label-list items.
This commit fixes the issue by properly appending a newline after each
label-list item so that the output of ToMarkdown can be reparsed by
RDoc::Markdown:
```
md = <<~MD
apple
: a red fruit
banana
: a yellow fruit
MD
doc = RDoc::Markdown.parse(mdoc)
doc # => [doc: [list: NOTE [item: ["apple"]; [para: "a red fruit"]], [item: ["banana"]; [para: "a yellow fruit"]]]]
new_md = doc.accept(RDoc::Markup::ToMarkdown.new)
new_md # => "apple\n: a red fruit\n\nbanana\n: a yellow fruit\n\n"
new_doc = RDoc::Markdown.parse(new_md)
new_doc # => [doc: [list: NOTE [item: ["apple"]; [para: "a red fruit"]], [item: ["banana"]; [para: "a yellow fruit"]]]]
```
[1]: https://michelf.ca/projects/php-markdown/extra/#def-listhttps://github.com/ruby/rdoc/commit/c65266437c
(https://github.com/ruby/irb/pull/897)
Some features of irb do not work properly when using the old rdoc.
I have compared several major versions and found that it works as intended from 4.0.0.
This problem occurs when there is a Gemfile.lock is installed with the old rdoc.
I don't know why this Gemfile.lock installs an older rdoc than the ruby bundled rdoc,
but specifying the version in the gemspec will at least prevent the problem.
NOTE: ruby/irb#704 problem does not occur with this change.
The following is test code.
```
### Usage: ruby __FILE__.rb
# # input RDoc and Tab
# >> RDoc<Tab>
#
### Expect: Display document of RDoc
### Actual: <internal:marshal>:34:in `load': instance of RDoc::Constant needs to have method `marshal_load' (TypeError)
require "bundler/inline"
gemfile(true) do
source "https://rubygems.org"
git_source(:github) { |repo| "https://github.com/#{repo}.git" }
gem 'irb'
# gem 'rdoc', '~> 4.0.0'
gem 'rdoc', '~> 3.12.0'
end
require 'rdoc'
require 'irb'
IRB.start
```
https://github.com/ruby/irb/commit/1a1fbba020
In cfd7729ce7 we started using inline
caches for refinements. However, we weren't clearing inline caches when
defined on a reopened refinement module.
Fixes [Bug #20246]
This `st_table` is used to both mark and pin classes
defined from the C API. But `vm->mark_object_ary` already
does both much more efficiently.
Currently a Ruby process starts with 252 rooted classes,
which uses `7224B` in an `st_table` or `2016B` in an `RArray`.
So a baseline of 5kB saved, but since `mark_object_ary` is
preallocated with `1024` slots but only use `405` of them,
it's a net `7kB` save.
`vm->mark_object_ary` is also being refactored.
Prior to this changes, `mark_object_ary` was a regular `RArray`, but
since this allows for references to be moved, it was marked a second
time from `rb_vm_mark()` to pin these objects.
This has the detrimental effect of marking these references on every
minors even though it's a mostly append only list.
But using a custom TypedData we can save from having to mark
all the references on minor GC runs.
Addtionally, immediate values are now ignored and not appended
to `vm->mark_object_ary` as it's just wasted space.
Kevin Menard
Jean Boussier
Koichi ITO
Hartley McGuire
alpaca-tc
dependabot[bot]
Nobuyoshi Nakada
Hiroshi SHIBATA
Kevin Newton
tompng
John Hawthorn
git
Stan Lo
BurdetteLamar
Ufuk Kayserilioglu
Maxime Chevalier-Boisvert