On Sat, 7 Nov 2009, Joshua Kwan wrote:

> This patch (should be applied on top of threading-v3.diff.gz) fixes the
> configure logic for evaluating whether we have an appropriate threadsafe
> resolver. The changes to hostthre.c have not yet been done yet.

I'm a bit curious on your talk and work on thread-safe resolvers.

If a getaddrinfo() or a gethostbyname_r() implementation anywhere is unsafe,
we already should have problems since libcurl is claimed to be thread-safe
already when using those functions and it has used them for many years without
problems.

Can anyone mention a getaddrinfo() or a gethostbyname_r() implementation that
isn't thread-safe?

The only systems that are hard to detect - to my recollection of this - are
the ones that provide a thread-safe gethostbyname() since it isn't that easy
to make a distinction, But in libcurl we prefer gethostbyname_r() anyway and
beyond that we rely on specific knowledge about for example that Window's name
resolver is thread-safe still.

-- 
  / daniel.haxx.se
-------------------------------------------------------------------
List admin: http://cool.haxx.se/list/listinfo/curl-library
Etiquette:  http://curl.haxx.se/mail/etiquette.html