vollkom...@gmx.net wrote:
Thanks for pointing these out. I'm not sure where the IPv6 headers are on OS X 
or if auto-detection would work so I'm leaving --enable-ipv6 in just to be safe.


:) enable only turns on the auto-detection.
Scan the config.log for "if PF_INET6 is available" says yes to confirm if its detected.

I'll try to set up SSH for remote access on my box (behind NAT) in a bit. In the 
meantime, I'm attaching a "make" output with gcc 4.2.1 (Apple Inc. build 5564) 
that appears to produce more verbose output. Maybe that'd provide a clue to the issue 
that gcc 4.0.1 doesn't.


Okay! this is much more helpful. Key bit is:

In file included from ../../include/Array.h:40,
                 from ../../src/acl/Acl.h:38,
                 from ./time.h:38,
                 from /usr/include/sys/time.h:235,
                 from /usr/include/sys/select.h:148,
                 from ../../include/squid_types.h:74,
                 from ../../compat/compat.h:45,
                 from ../../include/config.h:58,
                 from Acl.cc:34:
../../include/fatal.h:9: error: expected initializer before 'PRINTF_FORMAT_ARG1'


Seems we have a problem with system includes including squid includes and nasty loop begins. With time.h at fault.

Squid does not even have a file called time.h to get confused with (carefully calling our file SquidTime.h to be sure of that), and I seriously doubt any system file would normally include a squid-pathed file.

I'm no more closer to knowing how or how to prevent, but at least we know whats happening now.

Is this a perfectly clean build attempt, with no patches other than maybe the ones I've given you during the tracing?



-------- Original-Nachricht --------
Oooh. I don't think its relevant, but some cleanup is needed there...

Drop:
 --with-ipv6-split-stack;  it was partially added for Windows XP but never
completed due to windows VCS issues. Squid does dual-stack nicely.

 --enable-ipv6;  It's auto-detected. Just ensure you have the proper
v6-capable kernel and headers present.

 --enable-dl-malloc; obsolete

 --enable-storeio=coss; COSS has too many bugs in Squid-3 to be considered
usable.


No change in the trace. This is really, really weird. Am I able to get
shell access to the box you build these on? I'd like to test the build a
bit and first-hand is best for that.


Amos
--
Please be using
  Current Stable Squid 2.7.STABLE6 or 3.0.STABLE14
  Current Beta Squid 3.1.0.7

Reply via email to