ioctl accepts int as request arguments on some platforms, but some
requests are more than INT_MAX, e.g., RNDGETENTCNT(0x80045200).
Passing (0x80045200 | (-1 << 32)) may work around the issue, but it may
not work on a platform where ioctl accepts unsigned long. So this
change uses NUM2LONG and then casts it to int.
NtSocketsInitialized behavior changed in e33b1690, requiring
a call to rb_w32_sysinit for starting Windows Sockets.
This commit removes NtSocketsInitialized entirely to avoid confusion.
Signed-off-by: Gabriel Nagy <gabriel.nagy@puppet.com>
In the grammar, all expressions are statements, but not all
statements are expressions. Some parts of the grammar accept
expressions and not other types of statements, which causes
similar looking code to parse differently due to operator
precedence.
Mostly from Dan0042 (Daniel DeLorme).
Fixes [Bug #16092]
For BasicObject, bind the Kernel respond_to? instance method to the
object and call it instead of calling the method directly.
Also, use bind_call(recv, ...) for better performance.
Fixes [Bug #16127]
* ext/openssl/ossl_asn1.c (Init_ossl_asn1): prefer
`rb_gc_register_mark_object`, which is better for constant
objects, over `rb_gc_register_address` for global/static
variables which can be re-assigned at runtime. [Bug #16196]
Suggested by ko1. rb_fatal requires GVL so just in case one lacks,
print that information and let the process die. As commented,
we cannot print the given messages on such situations.
Requested by ko1 that ability of calling rb_raise from anywhere
outside of GVL is "too much". Give up that part, move the GVL
aquisition routine into gc.c, and make our new gc_raise().
* ext/openssl/ossl_asn1.c (Init_ossl_asn1): register the static
variable to grab an internal object, before creating the object.
otherwise the just-created object could get collected during the
global variable list allocation. [Bug #16196]
* parse.y (struct local_vars): moved numbered parameter NODEs for
nesting check to separate per local variable scopes, as numbered
parameters should belong to local variable scopes. [Bug #16248]
Now that allocation routines like ALLOC_N() can raise exceptions
on integer overflows. This is a problem when the calling thread
has no GVL. Memory allocations has been allowed without it, but
can still fail.
Let's just relax rb_raise's restriction so that we can call it
with or without GVL. With GVL the behaviour is unchanged. With
no GVL, wait for it.
Also, integer overflows can theoretically occur during GC when
we expand the object space. We cannot do so much then. Call
rb_memerror and let that routine abort the process.
On Android, a signal handler that is not SIG_DFL is set by default for
SIGSEGV. Ruby's install_sighandler inserts Ruby's handler only when the
signal has no handler, so it does not insert Ruby's SEGV report handler,
which caused some test failures.
This changeset forces to install Ruby's handler for some fatal signals
(sigbus, sigsegv, and sigill). They keep the original handlers, and
call them when the interpreter receives the signals.
Just refactoring.
The name "rb_bug_context" is completely unclear for me.
(Can you see that "context" means "machine register context"?)
The context is available only when a fatal signal (sigbus, sigsegv, or
sigill) is received; in fact, the function is used only for fatal
signals. So, I think the name should be changed.
The three functions for fatal signals, sigbus, sigsegv, and sigill, are
a family. The definition of ruby_abort had interrupted them for no
reason. This change just moves the definition after the family.
A method call is often with `argc = 1` and `argv = &v` where v is a
VALUE, and some functions shift the arguments by `argc-1` and `argv+1`
(for example, rb_sym_proc_call). I'm unsure whether it is safe or not
to pass a pointer `argv+1` to memcpy with zero length, but Coverity Scan
complains it. So this attempts to suppress the warning by explicit
check of the length.
The declaration of local variable in loop, it will initialize local variable for each run of the loop with clang generated code.
So, it shouldn't declare the local variable in heavy loop.
Array#sum with float elements will be faster around 30%.
* Before
user system total real
3.320000 0.010000 3.330000 ( 3.336088)
* After
user system total real
2.590000 0.010000 2.600000 ( 2.602399)
* Test code
require 'benchmark'
Benchmark.bmbm do |x|
ary = []
10000.times { ary << Random.rand }
x.report do
50000.times do
ary.sum
end
end
end
This changeset is to kill future possibility of bugs similar to
CVE-2019-11932. The vulnerability occurs when reallocarray(3)
(which is a variant of realloc(3) and roughly resembles our
ruby_xmalloc2()) returns NULL. In our C API, ruby_xmalloc()
never returns NULL to raise NoMemoryError instead. ruby_xfree()
does not return NULL by definition. ruby_xrealloc() on the other
hand, _did_ return NULL, _and_ also raised sometimes. It is very
confusing. Let's not do that. x-series APIs shall raise on
error and shall not return NULL.
