On 11/23/11 06:40, Amit Kulkarni wrote:
Benjamin,
If you are willing to test patches, I will be willing to update R to
latest version with feedback from previous months. I don't grok R, and
I need somebody who can give feedback that R is working as it should.
please let me know, and I will do it soon.
Absolutely. I just realized that the version in ports is to old to use
ggplot2 and plyr. So I would really like to help to get a new version
in.
Ok, there's a version of R in www.github.com/jasperla/openbsd-wip/
I will place the updated version of R in the same place. Give me a few
days to collect everything together.
thanks for the testing offers! they are hard to come by :-)
I have been testing R as it is, and I have not been able to
get it to function without patching main.c as follows. If it
is not patched, and the conditional compilation below is
triggered, then R (invariably??) generates a warning message
"*failed to set* *alternate signal stack*"
upon the failure of a call to sigaltstack(), continues on its
merry way despite this ghastly error, and all manner of data
corruption ensues.
However, I do not understand the theory of alternate signal
stacks or the implications of making all the threads use the
same stack, and I am rather intimidated by the author's
comment "NB: this really isn't safe..." in code that nobody
has dared touch in years.
Something is very wrong here, and I would much appreciate
if anybody has any ideas how to not only link it correctly
with -pthread, but to get it to function correctly, too.
--- src/main/main.c.orig Sun Sep 21 15:05:26 2008
+++ src/main/main.c Fri Nov 25 11:14:44 2011
@@ -434,7 +434,7 @@
}
#endif
-#if defined(HAVE_SIGALTSTACK) && defined(HAVE_SIGACTION) &&
defined(HAVE_WORKING_SIGACTION) && defined(HAVE_SIGEMPTYSET)
+#if 0
/* NB: this really isn't safe, but suffices for experimentation for now.
In due course just set a flag and do this after the return. OTOH,
