Comment 17 for bug 929219

I have seen some reports where glibc crashes if nscd is not running, especially with chromium but also with ktorrent. This is not always reproduceable, so there's some kind of race or random memory corruption.

valgrind reports:
==16750== Process terminating with default action of signal 11 (SIGSEGV)
==16750== Access not within mapped region at address 0x17
==16750== at 0xC63DC1E: __nscd_get_mapping (in /lib64/libc-2.15.so)
==16750== by 0xC63DDE3: __nscd_get_map_ref (in /lib64/libc-2.15.so)
==16750== by 0xC63B2E1: nscd_gethst_r (in /lib64/libc-2.15.so)
==16750== by 0xC63BB96: __nscd_gethostbyname2_r (in /lib64/libc-2.15.so)
==16750== by 0xC6213F0: gethostbyname2_r@@GLIBC_2.2.5 (in /lib64/libc-2.15.so)
==16750== by 0xC5F5D8F: gaih_inet (in /lib64/libc-2.15.so)
==16750== by 0xC5F6ED2: getaddrinfo (in /lib64/libc-2.15.so)
==16750== by 0x140E7BD: net::SystemHostResolverProc(std::string const&, net::AddressFamily, int, net::AddressList*, int*) (in /usr/lib64/chromium/chromium)
==16750== by 0x1407C57: net::HostResolverImpl::Job::DoLookup(base::TimeTicks const&, unsigned int) (in /usr/lib64/chromium/chromium)
==16750== by 0x2B5B2FA: base::(anonymous namespace)::WorkerThread::ThreadMain() (in /usr/lib64/chromium/chromium)
==16750== by 0x11753C1: base::(anonymous namespace)::ThreadFunc(void*) (in /usr/lib64/chromium/chromium)
==16750== by 0x919FF65: start_thread (in /lib64/libpthread-2.15.so)

gdb gives a similar backtrace.

After reverting commit 3a2c02424d9824f5cdea4ebd32ff929b2b1f49c6, the problem does not appear anymore.

I can't reproduce it yet on my system - so no further information.

Reports:
https://bugzilla.novell.com/show_bug.cgi?id=741021
https://bbs.archlinux.org/viewtopic.php?id=133021