Previously evdns was using HOST_NAME_MAX, and define it to 255 *only* if
it not set, however it does set on linux:
$ egrep -r define.*HOST_NAME_MAX /usr/include/bits
/usr/include/bits/local_lim.h:#define HOST_NAME_MAX 64
/usr/include/bits/posix1_lim.h:#define _POSIX_HOST_NAME_MAX 255
/usr/include/bits/confname.h:#define _SC_HOST_NAME_MAX _SC_HOST_NAME_MAX
But 64 should be the limit of the host component, not for the whole
hostname, as also noted by @ploxiln
So use our own EVDNS_NAME_MAX const, which is set to 255.
Fixes: #1280
build report:
[1/2] Building C object CMakeFiles/regress.dir/test/regress_mbedtls.c.o
/src/le/libevent/test/regress_mbedtls.c:50: warning: "SSL_get_peer_certificate" redefined
50 | #define SSL_get_peer_certificate mbedtls_ssl_get_peer_cert
|
In file included from /src/le/libevent/test/regress.h:139,
from /src/le/libevent/test/regress_mbedtls.c:35:
/src/oss/openssl/include/openssl/ssl.h:1799: note: this is the location of the previous definition
1799 | # define SSL_get_peer_certificate SSL_get1_peer_certificate
|
[2/2] Linking C executable bin/regress
But this should be cleaned up.
Since OpenSSL 3.0 there is new OPENSSL_VERSION_MAJOR, but previous
releases does not have it.
So let's use plain old OPENSSL_VERSION_NUMBER to avoid more preprocessor
macros.
* Don't use deprecated functions when building against OpenSSL 3.0.
* Recognise that OpenSSL 3.0 can signal a dirty shutdown as a protocol.
error in addition to the expected IO error produced by OpenSSL 1.1.1
* Update regress_mbedtls.c for compatibility with OpenSSL 3
mingw has upgraded to Mbed-TLS 3.1.0, so all the mingw tests that need
Mbed-TLS currently don't work.
v2: add missing mbedtls/version.h into test/regress_mbedtls.c
v3: suppress #warning "Including compat-2.x.h is deprecated" for mbedtls/compat-2.x.h
Compiler report:
/home/runner/work/libevent/libevent/test/regress_util.c: In function ‘test_event_calloc’:
/home/runner/work/libevent/libevent/test/regress_util.c:1318:4: error: argument 2 value ‘9223372036854775815’ exceeds maximum object size 9223372036854775807 [-Werror=alloc-size-larger-than=]
p = mm_calloc(EV_SIZE_MAX/2, EV_SIZE_MAX/2 + 8);
In file included from /home/runner/work/libevent/libevent/test/../util-internal.h:37:0,
from /home/runner/work/libevent/libevent/test/regress_util.c:30:
/usr/include/stdlib.h:541:14: note: in a call to allocation function ‘calloc’ declared here
extern void *calloc (size_t __nmemb, size_t __size)
v2: clang on CI does not have this option
TLS 1.3 does not support renegotation - it has been removed from the
protocol. Disable TLS 1.3 when testing this, so we negotiate an older
protocol version
Currently the code uses long, but long does not always have the same
representation as a pointer, such as on 64-bit Windows where long is
only 32-bit due to its unususal LLP64 ABI, but also on CHERI, and thus
Arm's prototype Morello architecture, where C language pointers are
represented as hardware capabilities, which have bounds, permissions and
other metadata to enforce spatial memory safety. Both of these cases
warn when casting a long to a pointer (Windows due to long being shorter
and thus it being likely you've truncated the address, and CHERI due to
long not having any capability metadata like pointers and thus it being
likely you've stripped the metadata, with the resulting "null-derived"
capability destined to trap if dereferenced), and in both cases casting
to intptr_t as the intermediate type instead will get rid of those
warnings.
I've got an issue when stop evconnlistener not in the event_base_loop()
thread. evconnlistener_disable() acquired lev->lock, if the same time,
user callbacks is runing, the event thread released lock, after callback
finished, it try to aquire the lock again, I think this makes conflict:
Here is backtraces:
thread 1:
0 __lll_lock_wait (futex=futex@entry=0x555555559a60, private=0) at lowlevellock.c:52
1 0x00007ffff7f2a131 in __GI___pthread_mutex_lock (mutex=0x555555559a60) at ../nptl/pthread_mutex_lock.c:115
2 0x00007ffff7f424c9 in evthread_posix_lock (mode=0, lock_=0x555555559a60) at evthread_pthread.c:79
3 0x00007ffff7f7dc12 in listener_read_cb (fd=7, what=2, p=0x5555555599a0) at listener.c:439
4 0x00007ffff7f6d758 in event_persist_closure (base=0x555555559370, ev=0x5555555599d8) at event.c:1645
5 0x00007ffff7f6da60 in event_process_active_single_queue (base=0x555555559370, activeq=0x5555555597e0,
max_to_process=2147483647, endtime=0x0) at event.c:1704
6 0x00007ffff7f6e018 in event_process_active (base=0x555555559370) at event.c:1805
7 0x00007ffff7f6e92a in event_base_loop (base=0x555555559370, flags=0) at event.c:2047
8 0x0000555555555449 in main () at test_listen.c:67
thread 2:
0 futex_wait_cancelable (private=<optimized out>, expected=0, futex_word=0x555555559858) at ../sysdeps/nptl/futex-internal.h:183
1 __pthread_cond_wait_common (abstime=0x0, clockid=0, mutex=0x555555559800, cond=0x555555559830) at pthread_cond_wait.c:508
2 __pthread_cond_wait (cond=0x555555559830, mutex=0x555555559800) at pthread_cond_wait.c:638
3 0x00007ffff7f426f3 in evthread_posix_cond_wait (cond_=0x555555559830, lock_=0x555555559800, tv=0x0) at evthread_pthread.c:162
4 0x00007ffff7f70bc5 in event_del_nolock_ (ev=0x5555555599d8, blocking=2) at event.c:2934
5 0x00007ffff7f70748 in event_del_ (ev=0x5555555599d8, blocking=2) at event.c:2821
6 0x00007ffff7f707a1 in event_del (ev=0x5555555599d8) at event.c:2830
7 0x00007ffff7f7d76e in event_listener_disable (lev=0x5555555599a0) at listener.c:343
8 0x00007ffff7f7d6e5 in evconnlistener_disable (lev=0x5555555599a0) at listener.c:325
9 0x00005555555552c3 in disable_thread (arg=0x5555555599a0) at test_listen.c:27
10 0x00007ffff7f27609 in start_thread (arg=<optimized out>) at pthread_create.c:477
11 0x00007ffff7e4e293 in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:95
During testing on win32, util/getaddrinfo failed with NULL hint info
r = evutil_getaddrinfo("www.google.com", NULL, NULL, &ai);
throwing a critical heap exception when evutil_freeaddrinfo is called.
This is because of improper use of freeaddrinfo when nodes within the
ai structure are allocated using mm_malloc or mm_calloc
(EVUTIL_AI_LIBEVENT_ALLOCATED)
This adds the flag in apply_socktype_protocol_hack and walks the linked
list in evutil_freeaddrinfo removing linked list nodes that are custom
allocated before calling freeaddrinfo.
Otherwise it breaks upload-artifacts:
Run actions/upload-artifact@v1
Uploading artifact 'ubuntu-18.04-cmake-NONE-build' from '/home/runner/work/libevent/libevent/build' for run #346
Uploading 1095 files
File error 'No such device or address' when uploading file '/home/runner/work/libevent/libevent/build/foo'.
...
Error: No such device or address
Error: Exit code 1 returned from process: file name '/home/runner/runners/2.282.1/bin/Runner.PluginHost', arguments 'action "GitHub.Runner.Plugins.Artifact.PublishArtifact, Runner.Plugins"'.
This will also suppress:
../test/regress_bufferevent.c: In function ‘test_bufferevent_read_failed’: 1102
../test/regress_bufferevent.c:1395:2: warning: ignoring return value of ‘write’, declared with attribute warn_unused_result [-Wunused-result] 1103
write(data->pair[0], buf, strlen(buf));
There are no standard for encoding a unix socket in an url. nginx uses:
http://unix:/path/to/unix/socket:/httppath
The second colon is needed to delimit where the unix path ends and where
the rest of the url continues.
Signed-off-by: Sean Young <sean@mess.org>
Add new flag (DNS_CNAME_CALLBACK) for
evdns_base_resolve_ipv4()/evdns_base_resolve_ipv6().
If set, you will get one more callback with type == DNS_CNAME and CNAME
in addrs argument.
The default timeout tolerance is 50 ms,
which causes intermittent failure in many the
related tests in arm64 QEMU.
See: https://bugzilla.yoctoproject.org/show_bug.cgi?id=14163
(The root cause seems to be a heavy load)
Signed-off-by: Yi Fan Yu <yifan.yu@windriver.com>
Looks like a `splice` implementation was planned, but has clearly never
eventuated (the TODO comment is from ~12 years ago, in
8b5bd77415fb6634fadf08357676926fecf5f032). For now, it's probably better
to remove the unused code/correct the docs.
The conflict cast convertion between the return value of
bufferevent_read() and evbuffer_remove(), int(-1)->size_t(An undefined
maximum)
Add test case of bufferevent_read() should return 0 in case of
evbuffer_remove() returns -1
Fixes: #1132
I recently found that when the network status changed when calling
bufferevent_socket_connect_hostname (e.g. switching between several
WIFIs), all DNS servers would fail, and the timeout of probe would be
very long if there were many DNS requests. I want libevent to support
manual setting of MAX_PROBE_TIMEOUT and TIMEOUT_BACKOFF_FACTOR
So move hardcoded MAX_PROBE_TIMEOUT and TIMEOUT_BACKOFF_FACTOR into
struct, and allow changing them.
Was not noticed since it is reported only with optimization enabled, as
stated in gcc(1):
The effectiveness of some warnings depends on optimizations also
being enabled. For example -Wsuggest-final-types is more effective with
link-time optimization and -Wmaybe-uninitialized does not warn at all
unless optimization is enabled.
And interesting thing is that it is reported only for -O2, not for -O3,
that's why I did not catched it in both cmake env that I had:
- debug (it has -O0)
- release (it has -O3)
While autoconf has -O2.
