This email went to [email protected], which has no one from Argonne on it. Cc’ing [email protected]. Yes, this should be CPPFLAGS, not CFLAGS. In practice it shouldn’t matter, but we should try to get that cleaned up. — Pavan On Jun 11, 2014, at 2:06 PM, <[email protected]> <[email protected]> wrote:
On Jun 11, 2014, at 11:01 AM, Service Account <[email protected]> wrote:
commit be93caec6c6b9b230fbde9b1b296cde001fde934 Author: Antonio J. Pena <[email protected]> Date: Mon Jun 9 17:12:17 2014 -0500
Added support for thread-safe errno in Solaris
The _REENTRANT macro is required to be defined in order to get thread-safe "errno" values. Modern POSIX systems provide this by default when using -lpthreads, but there are still some (Solaris) that require an explicit define. Since we weren't compiling with that macro, we were getting incorrect "errno" values from the threads (mainly 0).
More details: since we use nonblocking sockets, our code was checking "errno" to not be EINPROGRESS whenever we got an error code from a connect/accept, in order to determine an actual error (a return value of -1 and errno of EINPROGRESS does not define an actual error but a "not yet" when using nonblocking sockets).
Interesting, I did not know about this Solaris "quirk".
This patch automatically gets this macro defined by configure whenever we are not building in MPI_THREAD_SINGLE mode only.
Fixes #2107
Signed-off-by: Ken Raffenetti <[email protected]>
diff --git a/configure.ac b/configure.ac index 936f6ce..96545b0 100644 --- a/configure.ac +++ b/configure.ac @@ -1352,6 +1352,13 @@ fi export MPICH_THREAD_LEVEL AC_DEFINE_UNQUOTED(MPICH_THREAD_LEVEL,$MPICH_THREAD_LEVEL,[Level of thread support selected at compile time])
+# If not in MPI_THREAD_SINGLE, we need -D_REENTRANT to get thread-safe "errno". +# Most POSIX systems provide this by default when using -lpthread, but we only use it in MPI_THREAD_MULTIPLE. +# Some systems (Solaris) still require an explicit define in any case. +if test "$MPICH_THREAD_LEVEL" != "MPI_THREAD_SINGLE"; then + PAC_APPEND_FLAG([-D_REENTRANT], [CFLAGS])
A largely academic observation, but: should this be CPPFLAGS instead?
-Dave