On Wed, 18 Aug 2004, Nigel Horne wrote:

> Is Curl_resolve thread safe()?

Curl_resolv() (it is spelled incorrectly) is a wrapper function for a wide
range of different name resolvers (http://curl.haxx.se/docs/faq.html#5.9).
Stating anything on thread-safety generally for Curl_resolv() would be
strange.

> Valgrind reports errors such as this on every call. The unititialised value
> is of course more serious than the mutex is not locked error.

...

> ==29426== Use of uninitialised value of size 4
> ==29426== at 0x5489D6: gethostbyaddr_r@@GLIBC_2.1.2 (in
> /lib/libc-2.3.2.so)
> ==29426== by 0x51B77A: gaih_inet (in /lib/libc-2.3.2.so)
> ==29426== by 0x51C923: __GI_getaddrinfo (in /lib/libc-2.3.2.so)
> ==29426== by 0xF16CA6: (within /usr/lib/libcurl.so.2.0.2)

I believe this is a problem in (your version of) glibc alternatively a flaw in
valgrind. I can't see how libcurl is using getaddrinfo() the wrong way that
would prevent it from being thread-safe.

I don't get those problems on my Linux ipv6-enabled glibc 2.2.4 when I run the
curl tests.

-- 
      Daniel Stenberg -- http://curl.haxx.se -- http://daniel.haxx.se
       Dedicated custom curl help for hire: http://haxx.se/curl.html