linux: work around EOWNERDEAD io_uring kernel bug (#4002)

io_uring sometimes erroneously returns EOWNERDEAD when the intention was
to return 0. It's harmless and fixed in linux 5.14 so j

linux: work around EOWNERDEAD io_uring kernel bug (#4002)

io_uring sometimes erroneously returns EOWNERDEAD when the intention was
to return 0. It's harmless and fixed in linux 5.14 so just ignore the
error.

Fixes: https://github.com/libuv/libuv/issues/4001
Refs: https://github.com/torvalds/linux/commit/21f965221e

show more ...

freebsd: hard-code UV_ENODATA definition (#3999)

`FreeBSD` defines `ENODATA` in /usr/include/c++/v1/errno.h which is only visible
if C++ is the compilation unit. This can cause interop i

freebsd: hard-code UV_ENODATA definition (#3999)

`FreeBSD` defines `ENODATA` in /usr/include/c++/v1/errno.h which is only visible
if C++ is the compilation unit. This can cause interop issues when
integrating libuv build with C on a C++ project. Avoid this issue by
directly defining with the value defined in the file previously
mentioned.

show more ...

win: define MiniDumpWithAvxXStateContext always (#3998)

test: fix -Wbool-compare compiler warning (#3996)

Fixes: https://github.com/libuv/libuv/issues/3995

win: fix mingw build (#3994)

Fixes: https://github.com/libuv/libuv/issues/3992

unix: remove clang compiler warning pragmas (#3989)

Instead of suppressing the VLA warning, use a fixed-size buffer that is
big enough to receive at least one control message but not so

unix: remove clang compiler warning pragmas (#3989)

Instead of suppressing the VLA warning, use a fixed-size buffer that is
big enough to receive at least one control message but not so big that
IBMi PASE rejects it.

show more ...

unix: handle CQ overflow in iou ring (#3991)

When there are more than 128 concurrent cq completions the CQ ring
overflows as signaled via the `UV__IORING_SQ_CQ_OVERFLOW`. If this
hap

unix: handle CQ overflow in iou ring (#3991)

When there are more than 128 concurrent cq completions the CQ ring
overflows as signaled via the `UV__IORING_SQ_CQ_OVERFLOW`. If this
happens we have to enter the kernel to get the remaining items.

show more ...

unix: constrained_memory should return UINT64_MAX (#3753)

Document that we return UINT64_MAX if the cgroup limit is set to the
max. For cgroupv2, that happens if we encounter `max`, whil

unix: constrained_memory should return UINT64_MAX (#3753)

Document that we return UINT64_MAX if the cgroup limit is set to the
max. For cgroupv2, that happens if we encounter `max`, while cgroupv1
returns 9223372036854771712 when no limit is set (which according to
[this StackExchange discussion] is derived from LONG_MAX and
PAGE_SIZE). So make sure we also detect this case for cgroupv1.

[this StackExchange discussion]: https://unix.stackexchange.com/questions/420906/what-is-the-value-for-the-cgroups-limit-in-bytes-if-the-memory-is-not-restricte

Addresses: https://github.com/libuv/libuv/pull/3744/files#r974062912

show more ...

win,process: write minidumps when sending SIGQUIT (#3840)

This commit adds the ability to dump core when sending the `SIGQUIT`
signal on Windows. The change reads in the current registry

win,process: write minidumps when sending SIGQUIT (#3840)

This commit adds the ability to dump core when sending the `SIGQUIT`
signal on Windows. The change reads in the current registry setting for
local dumps, and attempts to write out to that location before killing
the process. See [collecting-user-mode-dumps] for registry and pathing
details. This behavior mimics that of the dumps created by the typical
Windows Error Reporting mechanism.

[collecting-user-mode-dumps]: https://learn.microsoft.com/en-us/windows/win32/wer/collecting-user-mode-dumps

show more ...

build: export compile_commands.json (#3968)

Allows LSP's like clangd to work.

test: add additional assert (#3983)

Add a missing assert that checks whether `uv_pipe_init()`
succeeded or not.

unix: use memcpy() instead of type punning (#3990)

Libuv makes no claim to being strict aliasing-clean but type punning
makes me feel unclean so replace it with memcpy().

docs: fix some typos (#3984)

macos: revert clonefile (#3987)

* Revert "macos: fix source not being followed when cloning (#3941)"

This reverts commit 507f2f950d5da7ab1586a46e81fc83dae10c0138.

* Revert

macos: revert clonefile (#3987)

* Revert "macos: fix source not being followed when cloning (#3941)"

This reverts commit 507f2f950d5da7ab1586a46e81fc83dae10c0138.

* Revert "darwin: bring back macos-specific copyfile(3) (#3654)"

This reverts commit d4eb276eea7cb19a888fe97d7759d97c7092ad02.

show more ...

macos: update minimum supported version (#3982)

Only support the versions still maintained by Apple.

linux: use io_uring to batch epoll_ctl calls (#3979)

This work was sponsored by ISC, the Internet Systems Consortium.

linux: fix logic bug in sqe ring space check (#3980)

Handle wraparound properly, otherwise we may end up overwriting elements
that have not been consumed by the kernel yet.

win: fix race condition in uv__init_console() (#3973)

Co-authored-by: Arlie Davis <ardavis@microsoft.com>
Fixes: https://github.com/libuv/libuv/issues/3970

build: remove freebsd and openbsd buildbots (#3974)

The buildbots are super flaky compared to other platforms and no one
seems invested enough to investigate. They don't add nearly enoug

build: remove freebsd and openbsd buildbots (#3974)

The buildbots are super flaky compared to other platforms and no one
seems invested enough to investigate. They don't add nearly enough value
for the friction they cause so away they go.

Fixes: https://github.com/libuv/libuv/issues/3934
Fixes: https://github.com/libuv/libuv/issues/3972
Refs: https://github.com/libuv/libuv/pull/3548

show more ...

unix,win: add uv_clock_gettime() (#3971)

Fixes: https://github.com/libuv/libuv/issues/1674

src: fix events/events_waiting metrics counter (#3957)

The worker pool calls all callbacks locally within the queue. So the
value of nevents doesn't properly reflect that case. Increase

src: fix events/events_waiting metrics counter (#3957)

The worker pool calls all callbacks locally within the queue. So the
value of nevents doesn't properly reflect that case. Increase the number
of events directly from the worker pool's callback to correct this.

In order to properly determine if the events_waiting counter needs to be
incremented, store the timeout value at the time the event provider was
called.

show more ...

test: add ASSERT_OK (#3957)

The majority of uses for ASSERT_EQ are to check if the return value is
0. So make a macro specifically for this and make things easier to read.

doc: update active maintainers list (#3967)

Move Bert and Fedor to emeriti.

macos: fix source not being followed when cloning (#3941)

Fixes: https://github.com/libuv/libuv/issues/3940

linux: remove bug workaround for obsolete kernels (#3965)

Libuv no longer supports such kernels so the workaround can be removed.