Comment 1 by kballard@apple.com:

I accidentally filed this with the wrong account. Is there any way to edit it to use my
customary account (kballard@gmail.com)?

Comment 2:

You would have to file it again, but you don't need to as this issue has already been
reported.
Please add any additional context to issue #1657 with your gmail account.

Status changed to Duplicate.

Merged into issue #1657.

Comment 3 by kballard:

Is this really the same issue? issue #1657 appears to be a problem that arises in the
test phase. My problem arises when trying to build the net package.

Comment 4:

undup'ed.

Status changed to Accepted.

Comment 5:

It looks like cgo assumes const unless it finds other information.
So is AI_MASK even in your system, and can a C program
that does #include <netdb.h> access it?

Comment 6 by kballard:

In /usr/include/netdb.h I find
#define AI_MASK \
    (AI_PASSIVE | AI_CANONNAME | AI_NUMERICHOST | AI_NUMERICSERV | \
    AI_ADDRCONFIG)

Comment 7 by kballard:

If it makes a difference, this is inside an #if block
#if !defined(_POSIX_C_SOURCE) || defined(_DARWIN_C_SOURCE)

Comment 8:

That doesn't necessarily mean anything.
If you write
#include <netdb.h>
int main() {
    return AI_MASK;
}
does it compile?

Comment 9 by kballard:

I take it back. I just looked up that value on the wrong system, still running 10.6.
Similarly, my comment about one system actually compiling above, that was this system.
That part should be completely ignored. And I obviously need more sleep.

Comment 10 by kballard:

Ok, now I've looked in /usr/include/netdb.h on an actual 10.7 system, and I find the
exact same thing.
/* valid flags for addrinfo (not a standard def, apps should not use it) */
#if !defined(_POSIX_C_SOURCE) || defined(_DARWIN_C_SOURCE)
#define AI_MASK \
    (AI_PASSIVE | AI_CANONNAME | AI_NUMERICHOST | AI_NUMERICSERV | \
    AI_ADDRCONFIG)
#endif /* (!_POSIX_C_SOURCE || _DARWIN_C_SOURCE) */

Comment 11:

Never try to read include files.  Only compile code.

Comment 12 by kballard:

I compiled the following program with clang:
#include <stdlib.h>
#include <netdb.h>
#include <stdio.h>
int main() {
    printf("AI_MASK: %d\n", AI_MASK);
    return 0;
}
and got as output
AI_MASK: 5127

Comment 13 by kwal@google.com:

I'm running the same build of OS X at home.
As far as I can tell, the version of GCC that ships with the Lion dev tools either
doesn't produce any EnumerationType DWARF entries, or is producing them in a way that
debug/dwarf doesn't recognize.  The dwarf.TagEnumerationType case in loadDWARF in
cgo/gcc.go doesn't ever seem to get run.  Without seeing these tags, cgo can determine
the value for simple const defines (e.g. #define FOO 23) but not more complicated
defines like AI_MASK.

Comment 14 by dmwright:

FYI, I'm seeing exactly this issue on OS X 10.6.7  I've attached my go/src/pkg/net as
well.
I compiled and ran the same test program as Kevin above, and got the same result:
AI_MASK: 5127

Attachments:

1. net.tgz (69481 bytes)

Comment 15 by briandorsey:

Compiling go worked fine for me a few weeks ago, but crashes with this same error now.
I've recently updated Xcode to the 4C104 build, version 4.2, released on June 6. I'm
running OS/X 10.6.7. It seems likely that something changed in this release of Xcode. 
I also ran the AI_MASK test code:
spark:golang$ cat test.c 
#include <stdlib.h>
#include <netdb.h>
#include <stdio.h>
int main() {
    printf("AI_MASK: %d\n", AI_MASK);
    return 0;
}
spark:golang$ clang test.c -o test.o
spark:golang$ ./test.o 
AI_MASK: 5127
I don't have a sense for where to start troubleshooting, but I'd be happy to try any
experiments which would help diagnose the issue.

Comment 16:

Reproduced with new Xcode on Snow Leopard.
Thanks for the tip.

Comment 17:

This issue was closed by revision 9968090.

Status changed to Fixed.

Comment 18:

This issue was closed by revision d292bc788668.