Hi, Some *BSD don't provide a reentrant gethostbyname: <q source='manpage of a BSD'> BUGS These functions use static data storage; if the data is needed for future use, it should be copied before any subsequent calls overwrite it. Only the Internet address format is currently understood. </q> The only one I know that doesn't work is openBSD >= 3.3. (produces random lookups) Does anyone know the situation better? It is for disabling threaded dns lookups at configure time. Cheers Jorge.- PS: That's it. A thread follows with other minutiae... ---------- Forwarded message ---------- Date: Sat, 13 Dec 2003 14:26:58 +0100 (MET) From: Anders Gavare <md1gavan@mdstud.chalmers.se> To: Jorge Arellano Cid <jcid@dillo.org> Subject: Re: [Dillo-dev] Re: BUG#494 On Sat, 13 Dec 2003, Jorge Arellano Cid wrote:
On Sat, 13 Dec 2003, Anders Gavare wrote:
How typical: now I found this already mentioned in dillo's README file. Sorry to have bothered you. :-) Still, this is something that should be detected and taken care of by the configure script...
Thanks for the info.
Can you send me a patch for configure.in?
Unfortunately I don't think I'm familiar enough with neither Dillo nor the configure system to add such a functionality. It took several days for me to track down the widget bug (#494), and several hours of testing to figure out this gethostbyname() bug.
Or at least the contents of the detected $target variable on your machine (an extra line with "echo $target" after AC_CANONICAL_SYSTEM will tell).
sparc64-unknown-openbsd3.3 The same problem should exist in other BSDs as well, but I've not actually seen the symptoms. The manpages of FreeBSD 4.5 and NetBSD-current (after 1.6.1) have the following in their gethostbyname(3) manpages: BUGS These functions use static data storage; if the data is needed for future use, it should be copied before any subsequent calls overwrite it. Only the Internet address format is currently understood. Imho, if Dillo aims for platform independance, which is one of the things I like best about Dillo, this feature (multithreaded DNS lookups) would be better off disabled _unless_ the underlying platform explicitly allows it. Not the other way around. That's my opinion, anyway. Anders