28375 - PR_NETDB_BUF_SIZE not resolveable on AIX-4.3.2

Status: RESOLVED FIXED

(SeaMonkey :: Build Config, defect, P3)

Description

[Ciaran Deignan]

In nsprpub/pr/include/prnetdb.h, line 64, PR_NETDB_BUF_SIZE is defined as
#if (defined(AIX) && defined(_THREAD_SAFE)) || defined(OSF1)
#define PR_NETDB_BUF_SIZE sizeof(struct protoent_data)
#else
#define PR_NETDB_BUF_SIZE 1024
#endif

However when this is used in netwerk/dns/public/_xpidlgen/nsIDNSListener.h,
struct protoent_data is not defined. (struct protoent_data is defined in
/usr/include/netdb.h). I added '#include <netdh.h>' in prnetdb.h, however
that wasn't enough for some unkonwn reason...

The error message generated is:
xlC_r -o nsSocketTransport.o -c -DAIX -DAIX4_3 -g -DMOZILLA_CLIENT -DNSCAP_DISAB
... -I/usr/local/include  -qflag=w:w -DPIC nsSocketTransport.cpp
"../../../dist/include/nsCppSharedAllocator.h", line 90.16: 1540-252: (W) The 
destructor for "unsigned short" does not exist. The call is ignored.
"../../../dist/include/nsIDNSListener.h", line 14.24: 1540-106: (S) The size of 
"protoent_data" is unknown in "sizeof" expression.

Comment 1

[Jim Dunn]

I am not seeing this on my build of today.

What I did find is that:
nsSocketTransport.cpp includes nsIDNSListener.h
nsIDNSListener.h includes prnetdb.h
prnetdb.h include prio.h
prio.h includes prinet.h
prinet.h includes <netdb.h>
<netdb.h> declares protoent_data

one thing, are you pulling NSPR yourself or are you allowing 
client.mk to pull it for you?

I noticed that you have MOZILLA_CLIENT defined.  I believe this
will compile NSPR with NSPR threads instead of posix threads,
are you doing this for a reason?  

Comment 2

[Ciaran Deignan]

I downloaded a new nightly build, removed itools-R6.3 from the PATH (to go
back to the AIX standard makedepend), tidied up a few environmental variables
(SOCKS_BANNER, DB2INSTANCE, DOCUMENT_DIRECTORY) and rebuilt from scratch.
It hasn't finished yet, but it appears to have successfully compiled 
nsSocketTransport.cpp.

I don't know where the -DMOZILLA_CLIENT is coming from, there is nothing in
the environment that looks related, and I'm using the standard command-line

# make -f client.mk pull_and_build_all
(I got CVS working)

I have made no modifications to the client.mk file. I had planned to upload
a complete make log, but I deleted it when I had to restart the make (disk 
full...)

Thanks for your patience,
Ciaran