Re: net-snmp not building on latest current 9.99.106

[Brad Spencer <brad%anduin.eldar.org@localhost>]

Dave Tyson <dtyson%anduin.org.uk@localhost> writes:

> There has been some collateral damage inflicted due to the merge of structures 
> in netinet6.h with netinet.h in recent NetBSD-current
>

[snip]

> in6pcb (now aliased to inpcb) no longer has in6p_af, in6p_laddr n6etc.
>
> Not sure how to workaround this. There are a shed load of defines dealing with 
> different versions of NetBSD and I guess a quick hack would be to check for 
> version 999010400 and set defines for in6p_af = inp_af etc. Does this seem 
> sensible? I could try and make a patch....

I noticed this fallout too.  Apparently there are sysctls available to
get the PCB information without groveling though kvm.

This is a bit of a snip of email exchange I had with Ryota Ozaki
<ozaki-r%netbsd.org@localhost> about net-snmp:

-> If a userland program uses sysctl (e.g., net.inet.tcp.pcblist) to get PCB data,
-> the program shouldn't be affected.  OTOH, if a userland program uses kvm(3)
-> (net-snmp does), the program needs to be changed to follow the kernel changes.
-> 
-> IIUC, the sysctl API keeps backward compatibility while kvm(3) doesn't.  So
-> I think it is better to make net-snmp use sysctl API if possible.

Although I use the net-snmp package quite a bit, I really don't have a
great personal opinion on what would be the best fix for net-snmp.
Assuming the work on making the NetBSD network stack MP continues more
changes could be coming.


> Dave





-- 
Brad Spencer - brad%anduin.eldar.org@localhost - KC8VKS - http://anduin.eldar.org