Subject: Re: pkg/31170: firefox package build dumps core
To: , <pkg-manager@netbsd.org>
From: David H.Gutteridge <dhgutteridge@sympatico.ca>
List: pkgsrc-bugs
Date: 09/07/2005 20:55:23
I found patch-br which specifically references this issue.  When
I originally looked at it (after I grepped pkgsrc for the version
issue in general last week), my brain was a bit fried and I misread
the "<" as a ">" (since ">" was what was typically coded for in
those version checks).  So I thought it was turning on behaviour
rather than turning it off.  (I remember thinking, "why LOCK_DNS
if the functions are re-entrant?" but it didn't occur to me to 
think about it for a minute and realize I was looking at it
backwards.)

So I've modified the patch to look for < 299000000 and am
recompiling it now, will report back when I'm done, I'm sure this
is the answer to the problem.

(I could smack myself for not really looking at the code.)

> 
> From: grant beattie <grant@NetBSD.org>
> Date: 2005/09/07 Wed PM 08:29:52 EST
> To: "David H. Gutteridge" <dhgutteridge@sympatico.ca>
> CC: tls@netbsd.org,  pkg-manager@netbsd.org,  gnats-admin@netbsd.org, 
> 	pkgsrc-bugs@netbsd.org
> Subject: Re: pkg/31170: firefox package build dumps core
> 
> On Wed, Sep 07, 2005 at 02:36:45PM -0400, David H. Gutteridge wrote:
> 
> > Hello fellow Firefox build sufferer,
> > 
> > I've successfully built Firefox several times on 2.1RC3/i386, several times
> > because if you do get it running, you may find it has what appear to be
> > problems relating to netdb functions.  I've put some emails out about the
> > issue, but don't know of a solution at this point.  (See for instance
> > http://mail-index.netbsd.org/current-users/2005/09/05/0003.html)
> > 
> > Basically, DNS queries hang, and they can do so even when the browser just
> > starts up and calls the home page that's set.  This appears to be related
> > to the fact that the pertinent function(s) in netdb.h are not thread-safe,
> > but I'm a bit confused as to why the exact same package built with the very
> > same options under 2.0.2/i386 works without such problems, I haven't figured
> > that out.  (So I may be following a red herring.)
> > 
> > I have two Firefox builds in package form, the plain vanilla version and one
> > with debugging symbols and output enabled.  If you're willing to trust some
> > guy you don't know on the internet (ha ha), I can make them available to you.
> 
> "ugh".
> 
> I think this is a fallout from the __NetBSD_Version__ change in 2.1,
> because firefox/mozilla/thunderbird test __NetBSD_Version__ and use
> thread-safe resolver functions if it's recent enough.
> 
> my guess is that the change in NetBSD 2.1 means this code is turned on
> in Firefox, but I think 2.1 lacks the thread-safe resolver functions,
> so it breaks.
> 
> look at patches/* and see if you can tweak the version number to do
> the right thing..
> 
> grant.
>