When reference updating ObjectSpace.trace_object_allocations, we need to
check whether the object is valid or not because it does not mark the
object so the object may be dead. This can cause a segmentation fault
if the object is on a free heap page.
For example, the following script crashes:
require "objspace"
objs = []
ObjectSpace.trace_object_allocations do
1_000_000.times do
objs << Object.new
end
end
objs = nil
# Free pages that the objs were on
GC.start
# Run compaction and check that it doesn't crash
GC.compact
Prism was already disallowing arguments after block args, but in
parse.y, any comma after a block arg is a syntax error. This moves the
error handling into `PM_TOKEN_UAMPERSAND` where we can check if the
current type is `PM_TOKEN_COMMA`then raise an error. I've also updated
the tests to include the examplesfrom ruby/prism#3112.
Fixes: ruby/prism#3112754cf8eddc
We need to reinsert into the ST table when an object moves because it is
a numtable that hashes on the object address, so when an object moves we
need to reinsert it rather than just updating the key.
RDoc::RubyGemsHook
(https://github.com/ruby/rdoc/pull/1244)
Rubygems creates an instance of RDoc::RubygemsHook, sets `doc.force = overwrite`, then calls `doc.generate` the document.
RDoc::RubygemsHook needs attribute `:force` just like RDoc::RubyGemsHook.
01bdbcdd4d
Compilation of NODE_HASH in compile_shareable_literal_constant does not support hash that contains keyword splat.
If there is a keyword splat, fallback to default case.
This change includes the following updates:
- Added an environment variable `RUBY_TCP_NO_FAST_FALLBACK` to control enabling/disabling fast_fallback
- Updated documentation and man pages
- Revised the implementation of Socket.tcp_fast_fallback= and Socket.tcp_fast_fallback, which previously performed dynamic name resolution of constants and variables. As a result, the following performance improvements were achieved:
(Case of 1000 executions of `TCPSocket.new` to the local host)
Rehearsal -----------------------------------------
before 0.031462 0.147946 0.179408 ( 0.249279)
after 0.031164 0.146839 0.178003 ( 0.346935)
-------------------------------- total: 0.178003sec
user system total real
before 0.027584 0.138712 0.166296 ( 0.233356)
after 0.025953 0.127608 0.153561 ( 0.237971)
[Bug #20950]
ifunc proc has the ep allocated in the cfunc_proc_t which is the data of
the TypedData object. If an ifunc proc is duplicated, the ep points to
the ep of the source object. If the source object is freed, then the ep
of the duplicated object now points to a freed memory region. If we try
to use the ep we could crash.
For example, the following script crashes:
p = { a: 1 }.to_proc
100.times do
p = p.dup
GC.start
p.call
rescue ArgumentError
end
This commit changes ifunc proc to also duplicate the ep when it is duplicated.
* Make it loose coupling between RubyGems and RDoc
\### Problems
There are following problems because of tight coupling between RubyGems and RDoc.
1. If there are braking changes in RDoc, RubyGems is also broken.
2. When we maintain RDoc, we have to change RubyGems.
The reason why they are happened is that RubyGems creates documents about a gem with installing it.
Note that RubyGems uses functions of RDoc to create documents.
Specifically,
- Creating documents is executed by `rubygems/lib/rubygems/rdoc.rb`.
- `::RDoc::RubygemsHook` which is defined by RDoc is called by the file.
\### Solution
RubyGems has the plugin system.
If a gem includes `rubygems_plugin.rb`, RubyGems loads it.
RubyGems executes a process defined in it while installing gems, uninstalling gems or other events.
We can use the system to solve the problems.
The root cause is RubyGems directly references the class of RDoc.
We can remove the root cause by making RDoc RubyGems plugin.
Alternatively `rubygems_plugin.rb` creates documents about gems.
\### FAQ
Q1. Do we need to change codes of RubyGems?
A.
No, we don't.
This change keeps compatibility of API used from RubyGems.
Q2. Is it better to delete existing codes related to RDoc in RubyGems?
No, it isn't.
If we change codes of RubyGems,
we can't keep a compatibility.
Example:
If we delete codes that uses `RDoc::RubygemsHook` in `rubygems/lib/rubygems/rdoc.rb`,
documentations are not created with old RDoc.
Q3. When can we delete `rubygems/lib/rubygems/rdoc.rb`?
A.
We can delete it when all users use RDoc including `rubygems_plugin`.
Next ruby version is 3.4.
If it includes the RDoc including `rubygems_plugin`,
we can delete `rubygems/lib/rubygems/rdoc.rb` after ruby 3.3 is EOL.
Q4. Is it a breaking change that Rubygems creates documents with
rubygems_plugin not RDoc::RubygemsHook?
A.
No, it isn't.
If we simply implement this approach,
we move the implementation from `rdoc/lib/rdoc/rubygems_hook.rb` to
`rubygems_plugin.rb`.
This way can be breaking change.
It seems to be fine that we just need to delete `rdoc/rubygems_hook.rb` but it doesn't work.
It generates multiple documents.
`rubygems/lib/rubygems/rdoc.rb` has the following code.
```
begin
require "rdoc/rubygems_hook"
# ...
rescue LoadError
end
```
This code ignores RDoc related processes when `rdoc/rubygems_hook` can't be required.
But, this 'require' is not failed.
This is because Ruby installs Rdoc as a default gem.
So, Rdoc installed as a default gem generates documents and one installed as a normal gem does it too.
If you think that this behavior is accectable,
we can just delete `rdoc/rubygems_hook.rb`.
What do you think about this approach?
In this change, we take another approach to solve the problem that creates multiple documents.
If `Gem.done_installing(&Gem::RDoc.method(:generation_hook))` in `rubygems/rdoc.rb` doesn't create documents,
we can solve the problem.
We have some options.
* We change `rubygems/rdoc.rb` and then don't execute `Gem.done_installing`.
(This is a change for RubyGems.)
* We change `rdoc/rubygems_hook.rb` and then make `generation_hook` a no-op method.
(This is a change for RDoc.)
We choose the latter to avoid changing for RubyGems.
\### Test
\#### Preparation
Install Rdoc which including our changes by executing `rake install`.
❯ rake install
We confirmed that Rdoc which including our changes was installed.
❯ gem list | grep rdoc
rdoc (6.6.0, default: 6.4.0)
\#### Check point
We tested to check compatibility.
How to chack the compatibility?
We tested creating same documents by our RDoc and old RDoc with latest RubyGems.
We used following versions to test.
```
❯ ruby -v
ruby 3.1.0p0 (2021-12-25 revision fb4df44d16) [arm64-darwin22]
❯ gem list | grep rdoc
rdoc (default: 6.4.0)
❯ ruby -I rubygems/lib rubygems/exe/gem --version
3.5.14
```
Here is a result of test with old RDoc.
We can see that the document is created correctlly with `Parsing...` and `Done installing...`.
```
❯ ruby -I rubygems/lib rubygems/exe/gem install pkg-config
Successfully installed pkg-config-1.5.6
Parsing documentation for pkg-config-1.5.6
Done installing documentation for pkg-config after 0 seconds
1 gem installed
```
Here is a result of test with our RDoc.
We can see that the document is created correctlly with `Parsing...` and `Done installing...`.
```
❯ ruby -I rubygems/lib rubygems/exe/gem install pkg-config
Successfully installed pkg-config-1.5.6
Parsing documentation for pkg-config-1.5.6
Done installing documentation for pkg-config after 0 seconds
1 gem installed
```
As you can see we got the same results, our RDoc keeps compatibility.
* rename a test file
* Revert "rename a test file"
This reverts commit 70a144bf3fb8f2cc653972e858b5fed3747765d7.
* revert a test class name
* exclude `TestRDocRubyGemsHook` at job of ruby-core
* When `rubygems_plugin.rb` is not found, `test_rdoc_rubygems_hook.rb` is skipped.
* remove unnecessary whitespace
* add comment
* Add support for the case that RDoc is installed as a default gem
* Fix problems
Co-authored-by: mterada1228 <49284339+mterada1228@users.noreply.github.com>
* Simplify
* removed unused blank lines and revert test
* for rerun tests
* add comment for rubygems_plugin.rb
---------
Co-authored-by: Sutou Kouhei <kou@clear-code.com>
Co-authored-by: Sutou Kouhei <kou@cozmixng.org>
(https://github.com/ruby/irb/pull/1046)
* Load history when starting a direct debug session
When starting a debug session directly with RUBY_DEBUG_IRB_CONSOLE=1 and
`require 'debug'; debugger`, IRB's history wasn't loaded. This commit ensures
history is loaded in this case by calling `load_history` when configuring IRB
for the debugger.
Fixesruby/irb#975
* Update test/irb/test_history.rb
* Update lib/irb/debug.rb
---------
7f851b5353
Co-authored-by: Stan Lo <stan001212@gmail.com>
(https://github.com/ruby/reline/pull/793)
* Undo and redo should restore indentation
Undo and redo should not perform auto indentation. It should not change the indentation. Instead, it should restore previous indentation.
* Rename ivar undoing(undoing or redoing) to restoring
6355a6e0b2
Raise an exception when the same numbered param is used inside a child
block. For example, the following code should be a syntax error:
```ruby
-> { _1 + -> { _1 } }
```
Fixes https://github.com/ruby/prism/pull/3291d4fc441838
parse.y treats CRLF as a LF and basically "normalizes" them before
parsing. That means a string like `%\nfoo\r\n` is actually treated as
`%\nfoo\n` for the purposes of parsing. This happens on both the
opening side of the percent string as well as on the closing side. So
for example `%\r\nfoo\n` must be treated as `%\nfoo\n`.
To handle this in Prism, when we start a % string, we check if it starts
with `\r\n`, and then consider the terminator to actually be `\n`. Then
we check if there are `\r\n` as we lex the string and treat those as
`\n`, but only in the case the start was a `\n`.
Fixes: #3230
[Bug #20938]
e573ceaad6
Co-authored-by: John Hawthorn <jhawthorn@github.com>
Co-authored-by: eileencodes <eileencodes@gmail.com>
Co-authored-by: Kevin Newton <kddnewton@gmail.com>
omitted on Fedora 40 aarch64
(https://github.com/ruby/fiddle/pull/161)
I found working on Fedora 40 on aarch64 that
`test_dlopen_linker_script_input_linux` was omitted due to not finding
libncurses.so, since it is in /usr/lib64. This PR makes the glob a
little more portable.
444774c0c4
(https://github.com/ruby/rdoc/pull/1141)
There have been several document refactors in ruby/ruby that extract
individual methods/classes' documentation into separate files, like
ruby/ruby#6567
Because RI is not capable of rendering those references, RI users
are left with dramatically fewer documentation on those methods/classes.
This commit adds a new option `--expand-ref` (default: true) to expand
all the rdoc-ref targets at the end of the output.
9e2b28c6e3
(https://github.com/ruby/rdoc/pull/1209)
* Use the original `label` description list style
As a default for all description lists, the original "label" style is
more readable.
This is slightly different from the original `label` dl though:
* slightly increased left margin for `dd` (to 1em)
* removed right margin on `dd`
* removed `dt` bottom margin and `dd` top margin, to reduce the gap
between the term and its description (to only the standard line-height
gap).
* Add closing tags for description list terms
Without the closing tags, the dt elements contain whitespace after the
text. This normally isn't a big deal, but does mess some things up,
e.g: using `::after` with `content: ", "` in stylesheets.
* Restore float:left style for note lists
Unlike the original note list styles, this version sets the line-height
for all `dt` elements to be the same as the `p` elements contained
inside the `dd`, so that the second line has the same indentation as all
subsequent lines.
* Add commas between note list terms
9e69ea6d75
Apply SSL options set in DEFAULT_PARAMS without clearing existing
options.
It currently clears options in order to avoid setting one of the
options included in OpenSSL::SSL::OP_ALL unless explicitly specified,
namely OpenSSL::SSL::OP_DONT_INSERT_EMPTY_FRAGMENTS. Now that
OpenSSL::SSL::OP_ALL has been removed from SSLContext#initialize, it is
no longer necessary.
77c3db2d65
Respect the SSL options set by default by SSL_CTX() and by the
system-wide OpenSSL configuration file.
OpenSSL::SSL::SSLContext#initialize currently adds OpenSSL::SSL::OP_ALL
on top of the default SSL options. Let's stop doing it.
OpenSSL::SSL::OP_ALL is a set of options that changes OpenSSL's behavior
to workaround various TLS implementation bugs. Using it is considered
usually safe, but is not completely harmless.
00bec0d905
Check the ID_callback_state ivar after SSL_read() or SSL_write()
returns, similar to what ossl_start_ssl() does.
Previously, callbacks that can raise a Ruby exception were only called
from ossl_start_ssl(). This has changed in OpenSSL 1.1.1. Particularly,
the session_new_cb will be called whenever a client receives a
NewSessionTicket message, which can happen at any time during a TLS 1.3
connection.
aac9ce1304
If `Gem::Specification.reset` is used, but there are still unresolved
dependencies, RubyGems prints a warning. There are though, certain cases
where the situation will not cause any issues.
One such case is when the unresolved dependency does not restrict any
versions (>= 0) and there's a default gem matching it.
In this situation, it doesn't matter if Gem paths change, because
default gems are still activatable, so the dependency will be properly
activated if ever needed.
e5f8a3068e
