A few race conditions in the testsuite

[Howard Hinnant]

On Mar 11, 2006, at 5:52 PM, Paolo Carlini wrote:

> Hi Howard,
>
> you are analysis and proposed fix make sense, of course. Only, I  
> would like to know which is the target showing the problem (x86- 
> darwin?!?), thus, be reassured that *on that target* the fix  
> actually works in practice (because I cannot do that check anyway:  
> all the targets I have available don't show the problem). If you  
> can help me on that, then, I can double-check on linux that  
> everything is still fine there, and commit the patch. A ChangeLog  
> entry would be also appreciated.


Ah, the ChangeLog.  Details, details. :-)

Index: libstdc++-v3/ChangeLog
===================================================================
--- libstdc++-v3/ChangeLog      (revision 111965)
+++ libstdc++-v3/ChangeLog      (working copy)
@@ -1,3 +1,10 @@
+2006-03-11  Howard Hinnant  <hhinnant@apple.com>
+
+       Fixed testsuite race conditions in:
+       * testsuite/27_io/basic_filebuf/underflow/char/10097.cc
+       * testsuite/27_io/objects/char/9661-1.cc
+       * testsuite/27_io/objects/wchar_t/9661-1.cc
+
2006-03-10  Paolo Carlini  <pcarlini@suse.de>
         * include/tr1/cmath: Add atan2 and pow bits; add using  
declarations.

I detected this on ppc-darwin actually.  And it is sporadic.   
Sometimes passes, sometimes fails.  Due to an unrelated OS bug,  
sometimes the failure resulted in a hang, which is how it originally  
got detected in the first place.  To reliably detect this one you  
have to add a std::cout << "Pass\n" to the parent process and build  
it separately, and then try it several times (at least that was what  
I did).

-Howard