[netperf-talk] OK, this time I really mean it?-)

[Rick Jones]

Andrew Gallatin wrote:
> Rick Jones wrote:
> 
>> So, I believe I have the last Windows tweaks from my windows contacts, 
>> and I think we've gone long enough between releases of netperf that we 
>> have more than enough fixes piled-up so I would like to declare 
>> netperf 2.4.5.  To that end I need your help - please pull the current 
>> top-of-trunk and give it a try on your platform of choice and let me 
>> know if anything is wrong in the next couple days.
> 
> 
> Two very minor things:
> 
> 1) When building svn using Ubuntu 8.04 as the "autotools" host,
>    AND when building outside the tree, I get a problem with
>    netperf_version.h not being created.  Eg:
> 
> svn co <....>
> cd trunk
> mkdir build.804
> cd build.804
> ../configure
> make
 >
> gcc -DHAVE_CONFIG_H -I. -I/home/gallatin/netperf/2.2-svn/trunk/src -I.. 
>     -g -
> O2 -MT netsh.o -MD -MP -MF .deps/netsh.Tpo -c -o netsh.o 
> /home/gallatin/netperf/
> 2.2-svn/trunk/src/netsh.c
> /home/gallatin/netperf/2.2-svn/trunk/src/netsh.c: In function 
> 'scan_cmd_line':
> /home/gallatin/netperf/2.2-svn/trunk/src/netsh.c:929: error: 
> 'NETPERF_VERSION' u
> ndeclared (first use in this function)
> /home/gallatin/netperf/2.2-svn/trunk/src/netsh.c:929: error: (Each 
> undeclared identifier is reported only once
> /home/gallatin/netperf/2.2-svn/trunk/src/netsh.c:929: error: for each 
> function it appears in.)
> 
> Somehow, it just copied netperf_version.h.in, rather than
> actually doing what its supposed to.   I've never seen this
> bug with local projects, so I'm guessing its something in
> netperf, though I don't know what..  Have I mentioned how
> much I love autotools??

Any autotools experts out there?  That netperf_version.h stuff was to give 
netperf/netserver a -V option.

> 2) Netperf on older Solaris (like S10GA) whines:
> 
>   WARNING! getaddrinfo returned a protocol other than the
>   requested protocol.  Please contact your vendor for
>   a fix to this bug in getaddrinfo()
> 
> There's nothing I can do about it, I didn't write the OS,
> and I hate having my output cluttered with seemingly
> useless noise.  Can we please make these conditional
> on debugging?   Proposed patch attached.

It was deliberately annoying because after doing all the work to convert to 
getaddrinfo() I was annoyed when it didn't work correctly on some platforms :) 
Is the issue with getaddrinfo() addressed in a later Solaris release?  If so, and 
there is indeed a patch to older versions out there I'd be inclined to keep the 
message (and yes, I've been bitten by it myself).  If there is no patch at all 
then I'm willing to take the patch.

happy benchmarking,

rick jones

> 
> Thanks,
> 
> Drew
> 
> 
> ------------------------------------------------------------------------
> 
> Index: src/nettest_bsd.c
> ===================================================================
> --- src/nettest_bsd.c	(revision 301)
> +++ src/nettest_bsd.c	(working copy)
> @@ -647,6 +647,9 @@
>    int    change_info = 0;
>    static int change_warning_displayed = 0;
>  
> +  if (!debug)	
> +    change_warning_displayed = -1;
> +
>    int count = 0;
>    int error = 0;
>