You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
glibc until 2.10 has a bug in gethostbyaddr_r that assumes the buffer
argument is 8-byte aligned
(http://sources.redhat.com/ml/libc-alpha/2009-01/msg00000.html). gcc
seems to always provide such alignment for the call in
socketmodule.c:socket_gethostbyaddr(), but llvm-gcc (possibly only HEAD,
not 2.4) does not, which causes a segfault in test_socket.py. The llvm
bug investigating the problem is http://llvm.org/bugs/show_bug.cgi?id=3233.
The attached patch specifies the alignment so that llvm-gcc and unfixed
glibcs work together.
I'll commit this tomorrow if there are no objections.
Activity
jyasskin commentedon Jan 8, 2009
glibc until 2.10 has a bug in gethostbyaddr_r that assumes the buffer
argument is 8-byte aligned
(http://sources.redhat.com/ml/libc-alpha/2009-01/msg00000.html). gcc
seems to always provide such alignment for the call in
socketmodule.c:socket_gethostbyaddr(), but llvm-gcc (possibly only HEAD,
not 2.4) does not, which causes a segfault in test_socket.py. The llvm
bug investigating the problem is http://llvm.org/bugs/show_bug.cgi?id=3233.
The attached patch specifies the alignment so that llvm-gcc and unfixed
glibcs work together.
I'll commit this tomorrow if there are no objections.
jyasskin commentedon Jan 9, 2009
Committed as r68450.
rpetrov commentedon Jan 10, 2009
What about socket_gethostbyname_ex() ?
jyasskin commentedon Jan 10, 2009
socket_gethostbyname_ex() calls gethostbyname_r() rather than
gethostbyaddr_r(), and that appears not to have the bug.