[Bug libstdc++/15015] Thread safety / race condition bug in iostreams on Solaris 5.9

[harri dot pasanen at trema dot com]

------- Additional Comments From harri dot pasanen at trema dot com  2004-04-20 12:13 -------
I compiled and installed gcc 3.3.3, and I could not provoke the same crash.  
 
But over several runs, I did get it to crash sporadically at exit, always with the 
backtrace: 
 
(gdb) bt 
#0  0xff2b5900 in __deregister_frame_info_bases (begin=0xff361e04) 
    at ../../gcc-3.3.3/gcc/unwind-dw2-fde.c:183 
#1  0xff32676c in __do_global_dtors_aux () 
   from /usr/gnu/gcc33/lib/gcc-lib/sparc-sun-solaris2.9/3.3.3/libstdc++.so.5 
#2  0xff35d0b4 in _fini () 
   from /usr/gnu/gcc33/lib/gcc-lib/sparc-sun-solaris2.9/3.3.3/libstdc++.so.5 
 
$ /usr/gnu/gcc33/bin/g++ -v 
Reading specs from /usr/gnu/gcc33/lib/gcc-lib/sparc-sun-solaris2.9/3.3.3/specs 
Configured with: ../gcc-3.3.3/configure --prefix=/usr/gnu/gcc33 --with-gnu-as 
--with-cpu=supersparc --with-gnu-ld --with-ld=/usr/gnu/bin/ld --with-as=/usr/gnu/bin/as 
--enable-version-specific-runtime-libs --enable-long-long --enable-threads 
--enable-languages=c,c++ 
Thread model: posix 
gcc version 3.3.3 
 
But this seems like a diffent bug, meriting a new entry, unless this is already known. 
 
As to the iostreams race condition that seems to be fixed, would it be feasible to 
merge the version from 3.3.3 to 3.2.3, and how would I go about it?  
 
 

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15015