Ah, sorry I forgot to reply.
The rmdir seems good to me.
I've been doing further testing and with just the upstream patches I
occasionally get the test errors but not always so there seems to be something
there. In addition I have tried to resurrect my futex patch. With following
tweaks it compiles:
In sb-concurrency tests I reduced test-mailbox-producers-consumer test sleep to
0.001
In threads.pure.lisp, remove openbsd from fails-on list on "wait-on-semaphore
semaphore-notification :lp-1038034" as futexes are enabled.
With above I got following test failure:
Failure: threads.pure.lisp / WITHOUT-INTERRUPTS+CONDITION-WAIT
Expected failure: threads.pure.lisp / (WAIT-ON-SEMAPHORE
SEMAPHORE-NOTIFICATION LP-1038034)
Skipped (broken): compiler.impure.lisp / BUG-308921
I really don't have ideas how to proceed debugging the timing issues.
Heres the updated futex patch:
$OpenBSD$
add futex support
Index: src/runtime/bsd-os.c
--- src/runtime/bsd-os.c.orig
+++ src/runtime/bsd-os.c
@@ -86,6 +86,11 @@ static void dragonfly_init();
#ifdef LISP_FEATURE_X86
#include <machine/cpu.h>
#endif
+#if defined(LISP_FEATURE_SB_THREAD) && defined(LISP_FEATURE_SB_FUTEX) \
+ && !defined(LISP_FEATURE_SB_PTHREAD_FUTEX)
+#include <sys/time.h>
+#include <sys/futex.h>
+#endif
static void openbsd_init();
#endif
@@ -690,7 +695,45 @@ os_dlsym(void *handle, const char *symbol)
return ret;
}
+#if defined(LISP_FEATURE_SB_THREAD) && defined(LISP_FEATURE_SB_FUTEX) \
+ && !defined(LISP_FEATURE_SB_PTHREAD_FUTEX)
+
+int
+futex_wait(int *lock_word, int oldval, long sec, unsigned long usec)
+{
+ struct timespec timeout;
+ int ret;
+
+ if (sec < 0) {
+ ret = futex(lock_word, FUTEX_WAIT, oldval, NULL, NULL);
+ }
+ else {
+ timeout.tv_sec = sec;
+ timeout.tv_nsec = usec * 1000;
+ ret = futex(lock_word, FUTEX_WAIT, oldval, &timeout, NULL);
+ }
+
+ switch (ret) {
+ case 0:
+ return 0;
+ case ETIMEDOUT:
+ return 1;
+ case EINTR:
+ return 2;
+ default:
+ /* EWOULDBLOCK and others, need to check the lock */
+ return -1;
+ }
+}
+
+int
+futex_wake(int *lock_word, int n)
+{
+ return (futex(lock_word, FUTEX_WAKE, n, NULL, NULL));
+}
+
#endif
+#endif /* __OpenBSD__ */
#if defined(LISP_FEATURE_SB_WTIMER) && !defined(LISP_FEATURE_DARWIN)
/*
Josh Elsasser <[email protected]> writes:
> On Thu, Jan 02, 2020 at 08:14:10PM +0200, Timo Myyrä wrote:
>
>> Solene Rapenne <[email protected]> writes:
>>
>> > On Wed, Jan 01, 2020 at 07:00:45PM +0200, Timo Myyrä wrote:
>> >
>> >> Hi,
>> >>
>> >> Here's an attempt to update sbcl to latest version.
>> >> Before adding the new patch I got consistent test failures in inpure
>> >> timer test
>> >> "(:WITH-TIMEOUT :MANY-AT-THE-SAME-TIME".
>> >>
>> >> After adding the patch to unix.lisp the tests have passed twice on amd64.
>> >> Does this look OK, does the tests work on other platforms?
>> >>
>> >> Timo
>> >>
>> >
>> > I'm ok and will commit it if maintainer is ok too
>>
>> Heres revised diff with upstream backports. I still think we're better off
>> waiting next release so we don't have to include these patches but it can't
>> hurt
>> to test these to see they indeed fix the issue.
>>
>> Tested clisp bootstrapped on amd64 and tests passed.
>>
>> timo
>
> I sent a reply a few days ago but it's gone missing, so let's try again.
>
> The timing-related tests can be a bit fragile, I think the
> gettimeofday() change is only hiding a heisenbug in the tests for
> you. Personally, I think this is a test bug and not a real sbcl
> problem and isn't really worth patching.
>
> Additionally, you've added an empty html doc directory to the
> PLIST. How about this, which rmdir's the directory so it doesn't get
> re-added? I've tested and i386/macppc test results aren't worse.
>
> Index: Makefile
> ===================================================================
> RCS file: /cvs/ports/lang/sbcl/Makefile,v
> retrieving revision 1.43
> diff -u -r1.43 Makefile
> --- Makefile 16 Sep 2019 06:24:18 -0000 1.43
> +++ Makefile 2 Jan 2020 18:00:11 -0000
> @@ -6,7 +6,7 @@
>
> COMMENT= compiler and runtime system for ANSI Common Lisp
>
> -V = 1.5.5
> +V = 2.0.0
> DISTNAME= sbcl-${V}-source
> PKGNAME= sbcl-${V}
> WRKDIST= ${WRKDIR}/sbcl-${V}
> @@ -82,6 +82,7 @@
>
> post-install:
> chown -R 0:0 ${PREFIX}/lib/sbcl
> + rmdir ${PREFIX}/share/doc/sbcl/html
>
> do-test:
> cd ${WRKSRC} && ${SETENV} ${MAKE_ENV} DONT_CLEAN_SBCL_CONTRIB=1 \
> Index: distinfo
> ===================================================================
> RCS file: /cvs/ports/lang/sbcl/distinfo,v
> retrieving revision 1.18
> diff -u -r1.18 distinfo
> --- distinfo 16 Sep 2019 06:24:18 -0000 1.18
> +++ distinfo 2 Jan 2020 18:00:11 -0000
> @@ -1,2 +1,2 @@
> -SHA256 (sbcl-1.5.5-source.tar.bz2) =
> y0f65qhvDFxXQxYE+05fEcioI/lM4SjVaLh3D8W8quI=
> -SIZE (sbcl-1.5.5-source.tar.bz2) = 6351480
> +SHA256 (sbcl-2.0.0-source.tar.bz2) =
> kDaSVoBdQ3yCq5vauaQQB29XgQpQuysijeTmyJJpL88=
> +SIZE (sbcl-2.0.0-source.tar.bz2) = 6457217
