Am 2020-04-24 um 18:23 schrieb Christopher Schultz:
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256
Michael,
On 4/23/20 18:42, micha...@apache.org wrote:
This is an automated email from the ASF dual-hosted git
repository.
michaelo pushed a commit to branch master in repository
https://gitbox.apache.org/repos/asf/tomcat-native.git
The following commit(s) were added to refs/heads/master by this
push: new f95f531 Introduce tcn_get_thread_id(void) to reduce code
duplication f95f531 is described below
commit f95f531e98278cc7555367084b967e3550734559 Author: Michael
Osipov <micha...@apache.org> AuthorDate: Thu Apr 23 18:52:44 2020
+0200
Introduce tcn_get_thread_id(void) to reduce code duplication
At two spots (ssl.c and thread.c) we need to obtain the native
thread id. This has been done with two different approaches. Move
out to tcn_get_thread(void) which uses the previous
ssl_thread_id(void) implementation while the previous functions
delegate to the new one. apr_os_thread_current(void) is not used
anymore which does internally the same thing as
ssl_thread_id(void) was doing.
Also add properly #ifdefs for Windows and macOS for function
prototype includes. --- native/include/tcn.h | 1 +
native/src/jnilib.c | 45
+++++++++++++++++++++++++++++++++++++++ native/src/ssl.c
| 33 +--------------------------- native/src/thread.c
| 3 ++- xdocs/miscellaneous/changelog.xml | 5 ++++- 5 files
changed, 53 insertions(+), 34 deletions(-)
diff --git a/native/include/tcn.h b/native/include/tcn.h index
2b2ae59..d2f316b 100644 --- a/native/include/tcn.h +++
b/native/include/tcn.h @@ -175,6 +175,7 @@ char
*tcn_strdup(JNIEnv *, jstring); char *tcn_pstrdup(JNIEnv
*, jstring, apr_pool_t *); apr_status_t
tcn_load_finfo_class(JNIEnv *, jclass); apr_status_t
tcn_load_ainfo_class(JNIEnv *, jclass); +unsigned long
tcn_get_thread_id(void);
#define J2S(V) c##V #define J2L(V) p##V diff --git
a/native/src/jnilib.c b/native/src/jnilib.c index dae3ade..e88d4d5
100644 --- a/native/src/jnilib.c +++ b/native/src/jnilib.c @@ -23,6
+23,22 @@
#include "tcn_version.h"
+#ifdef WIN32 +#include <Windows.h> +#endif + +#ifdef DARWIN
+#include <pthread.h> +#endif + +#ifdef __FreeBSD__ +#include
<pthread_np.h> +#endif + +#ifdef __linux__ +#include
<sys/syscall.h> +#endif + #ifdef TCN_DO_STATISTICS extern void
sp_poll_dump_statistics(); extern void
sp_network_dump_statistics(); @@ -481,3 +497,32 @@ jint
tcn_get_java_env(JNIEnv **env) } return JNI_OK; } + +unsigned long
tcn_get_thread_id(void)
Why not simple call apr_os_thread_current() instead of writing a new
function? Or is the intention to get away from using APR?
There are several reasons here. Let me go in detail:
* Since we'd like to reduce the dependecy on libapr in the long run,
this is a small preparation.
* When you take a look at how apr_os_thread_current(void) is implemented
you will see that is is mediocre, on Unix(-like) it calls pthread_self().
Our approach for ssl_thread_id(void) is been better than the APR
counterpart. I have done a research why pthread_self() isn't enough and
an overwhelming amount of voices says that pthread_self() gives very
little information, but the thread pointer which is not the thread id.
So this is the last resort, but not a real solution.
+{ + /* OpenSSL needs this to return an unsigned long. On
OS/390, the pthread + * id is a structure twice that big. Use
the TCB pointer instead as a + * unique unsigned long. +
*/ +#ifdef __MVS__ + struct PSA { + char unmapped[540]; +
unsigned long PSATOLD; + } *psaptr = 0; + + return
psaptr->PSATOLD;
I think we might want to put the above #ifdef as the LAST one in the
list. I think if we can call pthread_self, we should, even if this
other technique will work.
I do not fully understand this. Should the test for z/OS happen before
the generic code? The code comes from here [1].
+#elif defined(WIN32) + return (unsigned
long)GetCurrentThreadId(); +#elif defined(DARWIN) + uint64_t
tid; + pthread_threadid_np(NULL, &tid); + return (unsigned
long)tid; +#elif defined(__FreeBSD__) + return (unsigned
long)pthread_getthreadid_np(); +#elif defined(__linux__) +
return (unsigned long)syscall(SYS_gettid); +#else + return
(unsigned long)pthread_self(); +#endif
Can we guarantee that pthread_self() will be available "by default"
since it's predicate-less, here? I wasn't able to quickly find a
reference for how to check for pthreads. Maybe #ifdef(_PTHREAD_H)?
That's a very valid point I forgot. I will investigate and test. Thank
you for raising! Surprisingly, I did not get a warning about a missing
prototype by HP aC.
M
[1] https://bz.apache.org/bugzilla/show_bug.cgi?id=56210
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org
For additional commands, e-mail: dev-h...@tomcat.apache.org