On Thu, 18 Sep 2008 22:47:22 +0200, Gennaidy Rozental wrote:

Boris writes:

...

I've a test case which receives the SIGCHLD signal on Solaris. Boost.Test catches it and prints an error message although in that test case SIGCHLD s perfectly fine. How do I make Boost.Test ignore this signal? I found

What error code does it produce?

child has exited; pid: 0; uid: 28570; exit value: 0

You can always use --catch_system_errors=no

I assume all signals are then ignored? As I'm also afraid of telling other developers again and again to use certain command line options I'd prefer to use a preprocessor directive.

In trunk I checked in compile time switch that affects handling child processes with non-zero return codes

I temporarily hardcoded in boost/test/impl/execution_monitor.ipp: m_CHLD_action( SIGCHLD, false, attach_dbg, alt_stack ) That works. If the macro BOOST_TEST_IGNORE_NON_ZERO_CHILD_CODE does the same it should be all right. :) Boris

On Fri, 19 Sep 2008 01:00:09 +0200, Gennadiy Rozental wrote:

Boris writes:

...

I temporarily hardcoded in boost/test/impl/execution_monitor.ipp:

m_CHLD_action( SIGCHLD, false, attach_dbg, alt_stack )

That works. If the macro BOOST_TEST_IGNORE_NON_ZERO_CHILD_CODE does the same it should be all right. :)

No. We still want to detect if child is crashed, for example.

I see. Can you change the code then to something like this: static bool ignore_sigchild( siginfo_t* info ) { return info->si_signo == SIGCHLD #ifdef BOOST_TEST_IGNORE_SIGCHLD ; #else && info->si_code == CLD_EXITED # ifdef BOOST_TEST_IGNORE_NON_ZERO_CHILD_CODE ; # else && (int)info->si_status == 0; # endif #endif } I have another test case where the parent kills the child. Simply ignoring SIGCHLD is probably the best (as otherwise there have to be too many preprocessor directives added)? Boris