| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| | |
|
| | |
|
| |\
| |
| |
| |
| | |
* rickard/busy-port-fix/21/GH-4898/OTP-17448:
Fix busy port handling in port_command()
|
| | |\
| | |
| | |
| | |
| | |
| | |
| | | |
rickard/busy-port-fix/21/GH-4898/OTP-17448
* rickard/busy-port-fix/GH-4898/OTP-17448:
Fix busy port handling in port_command()
|
| | | | |
|
| |\ \ \
| |/ /
|/| |
| | |
| | | |
* sverker/21/ets-matchspec-stack-bug/OTP-17379:
erts: Fix matchspec stack depth estimation bug
|
| | | | |
|
| | | | |
|
| | | | |
|
| |\ \ \
| | | |
| | | |
| | | |
| | | | |
* kjell/runtime_tools/dbg/ERL-1371/OTP-16930/GH-4396:
GH-4396: exported function instead of fun to set up remote tracing
|
| | |/ /
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
The function `dbg:n/1` used a local fun to set up a tracer on a remote
node. This works fine as long as the remote node is running exactly
the same version of Erlang/OTP but does not work at all
otherwise. This is fixed by exporting the relevant function
(`dbg:do_relay/2`) and by calling this function on the remote node to
set up remote tracing.
|
| |\ \ \
| | | |
| | | |
| | | |
| | | | |
* lukas/erl_interface/send-inf-tmo-fix/OTP-17358:
ei: Correctly check timeout value vs EI_SCLBK_INF_TMO
|
| | |/ /
| | |
| | |
| | |
| | |
| | | |
For the timeout (passed as `ms`) zero does not represent
infinity, but rather no sleeping at all. So that is what
we should check against and not 0.
|
| |\ \ \
| | | |
| | | |
| | | |
| | | | |
* rickard/off-heap-msgq-fix/OTP-17349:
Fix change of message_queue_data to off_heap
|
| | |/ / |
|
| |\ \ \
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
* rickard/signal-order-fix/21/OTP-17291:
Ensure is_process_alive() and process_info() respect signal order
Ensure reply signals from ports respect signal order
Ensure GC and CPC signals respect signal order
|
| | |\ \ \
| | | |/
| | |/|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
rickard/signal-order-fix/21/OTP-17291
* rickard/signal-order-fix/OTP-17291:
Ensure is_process_alive() and process_info() respect signal order
Ensure reply signals from ports respect signal order
Ensure GC and CPC signals respect signal order
|
| | | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
The reply signal sent as a result of a call to the is_process_alive() or
the process_info() BIFs could pass before other signals sent between the
involved processes. That is, the reply signals could be delivered too
early.
|
| | | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
The reply signal from a port to a process, when the process called one of the
'port_*' BIFs, could pass before other signals sent from the port to the
process. That is, the reply signal could be delivered too early.
|
| | | |/
| | |
| | |
| | |
| | |
| | | |
'garbage-collect' signals and 'check-process-code' signals could pass before
other signals sent from one process to another. That is, they could be
delivered too early.
|
| |\ \ \
| | | |
| | | |
| | | |
| | | | |
* rickard/literal-whole-message/OTP-17307:
Fix literal collection of messages
|
| | | |/
| |/|
| | |
| | |
| | | |
The literal GC missed copying literals in messages where the whole
message consisted of a literal.
|
| | | | |
|
| | | | |
|
| |\ \ \
| | | |
| | | |
| | | |
| | | |
| | | | |
* hans/ssh/idle_time-bug/OTP-17279:
ssh: Fix idle_time bug
ssh: Update idle_time test case
|
| | | | | |
|
| | | | | |
|
| |\ \ \ \
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
maint-21
* rickard/multizero-timeout-in-timeout-fix/21/OTP-17253:
erts: Fix bug in timer wheel at cancel during bump yield
erts: Tidy up some harmless code typos
Fix multi-zero timeout callback handling in timer wheel
|
| | | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Symptom:
Failed debug assert in find_next_timeout
ERTS_TW_ASSERT(tiw->yield_slot == ERTS_TW_SLOT_INACTIVE);
Problem:
If remove_timer() was called in between yielding erts_bump_timers()
tiw->true_next_timeout_time could be set to 0 leading to
find_next_timeout() being called before all bumping is done.
Solution:
Don't clear tiw->true_next_timeout_time in remove_timer()
if tiw->yield_slot is active.
Does not seem this bug could cause other more harmful symptoms,
but not sure.
|
| | | | | | |
|
| | |\ \ \ \
| | |_|_|/
| |/| | |
| | | | |
| | | | |
| | | | |
| | | | | |
rickard/multizero-timeout-in-timeout-fix/21/OTP-17253
* rickard/multizero-timeout-in-timeout-fix/OTP-17253:
Fix multi-zero timeout callback handling in timer wheel
|
| | | | |/
| | |/|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Execution of a zero timeout callback which was started from another
zero timeout callback could be delayed. Most timers in the system
do not behave like this. The service timer placed in the timer wheel
for management of the red-black tree of very long timers may however
behave like this.
|
| |\ \ \ \
| |/ / /
|/| | |
| | | |
| | | |
| | | |
| | | |
| | | | |
* lukas/erts/sendfile_active_close/OTP-17244:
kernel: Make sure to close file so that fd check work
erts: Return badarg if multiple sendfiles happen
kernel: Add fd leak checks to sendfile SUITE
erts: Fix fd leak when an active socket is closed remotely
|
| | | | | |
|
| | | | | |
|
| | | | | |
|
| | |/ /
| | |
| | |
| | | |
Fixes #4540
|
| | | | |
|
| | | | |
|
| |\ \ \
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
* bjorn/macos11-otp21/OTP-16779:
inets: Eliminate warnings in C code in tests
kernel: Eliminate warnings in C code in tests
erts: Eliminate warnings in C code in tests
Add missing include statements in in6addr tests
Update config scripts to compile for Apple Silicon (arm64)
macOS: Remove version check in configure script
Fix segfault(s) from variadic args on ARM64
Undo perform degradation on macOS
|
| | | | |
| | | |
| | | |
| | | |
| | | | |
Apple Clang 12 in Xcode 12 reports an error for the use of a
function without a declaration.
|
| | | | |
| | | |
| | | |
| | | |
| | | | |
Apple Clang 12 in Xcode 12 reports an error for the use of a
function without a declaration.
|
| | | | |
| | | |
| | | |
| | | |
| | | | |
Apple Clang 12 in Xcode 12 reports an error for the use of a
function without a declaration.
|
| | | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Compilation is failing on macOS Big Sur due to in6addr_any and in6addr_loopback
being redeclared. The underlying issue is that the tests for these are failing
due to missing include statements:
conftest.c:197:1: error: implicitly declaring library function 'printf'
with type 'int (const char *, ...)' [-Werror,-Wimplicit-function-declaration]
printf("%d", in6addr_any.s6_addr[16]);
^
conftest.c:197:1: note: include the header <stdio.h> or explicitly provide
a declaration for 'printf'
ERL-1306
|
| | | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Detect Apple Silicon (arm64) during configuration for compilation
on macOS 11.0 (beta at this time).
(This is a partial cherry-pick of b93b9f95438.)
|
| | | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
There is a version check in one of the configure scripts that
tests the version of macOS to ensure that the executables that
are built can be run on the same computer that built them.
The reason for this test has been lost in the mists of time.
Presumably there was a good reason for the test when it was
first added.
Since the test recently has become more complicated because of the
introduction of macOS 11.0 and 11.0.1, and is incorrect for some
versions of macOS, remove it. If there turns out that the version test
would be useful, it can be added back later (but in a less buggy way,
preferably using `awk` instead of `sed` to make the version parsing
simpler and more accurate).
|
| | | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Fix for Segfaults on Apple ARM64. Apple silicon does not support
re-casting functions with fixed params as variadic, and instead
requires the the function pointer to match the signature of the
definition. If they do not match arguments will not be correctly
marshaled, causing the function implementation to look for
parameters in the wrong places. To enable Apple ARM64 support, we
must explicitly cast to the correct function signature instead of
using varargs `...`.
Co-Authored-By: mjc <mjc@kernel.org>
|
| | | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Apple Clang 12 included in Xcode 12 reports an error when a
function without a declaration is used. That change causes
the configure test for jump table support to wrongly assume
that jump table support is missing, and therefore a `switch`
would be used, degrading the performance of BEAM.
|
| |\ \ \ \
| | | | |
| | | | |
| | | | |
| | | | | |
* rickard/aux-work-fix/OTP-17185:
Add missing memory barriers in aux work handling
|
| | | |/ /
| |/| | |
|
