In package "net-snmp-5.0.8" an include-file is corrupted, the file gets installed to /usr/include/net-snmp/library/data_list.h Here is a patch: diff -a -U 2 old/data_list.h new/data_list.h --- old/data_list.h 2003-08-09 12:02:30.000000000 +0000 +++ new/data_list.h 2003-08-09 12:01:49.000000000 +0000 @@ -27,5 +27,5 @@ - NETSNMP_INLINE netsnmp_data_list *netsnmp_create_data_list(const char *, + netsnmp_data_list *netsnmp_create_data_list(const char *, void *, Netsnmp_Free_List_Data @@ -39,5 +39,5 @@ int netsnmp_remove_list_node(netsnmp_data_list **realhead, const char *name); - NETSNMP_INLINE void *netsnmp_get_list_node(netsnmp_data_list *head, + void *netsnmp_get_list_node(netsnmp_data_list *head, const char *name);
This problem isn't what you think. The real problem is that gcc includes the following header: /usr/lib/gcc-lib/alphaev67-unknown-linux-gnu/3.2.3/include/net-snmp/net-snmp-config.h which overrides the right one: /usr/include/net-snmp/net-snmp-config.h This results in NET_SNMP_INLINE being undefined because the header in the gcc directories is old. The problem appears to be that gcc is sucking in the header file as part of fixincludes. This is not a net-snmp bug; it's a gcc bug. For the moment you can probably fix the problem by removing the file from the gcc include directory. But gcc needs to be fixed to prevent this problem.
I got bitten by a simular problem as this and I think I might have a solution to this problem. I think that we should run "fixincl.sh" from the gcc distribution after every installed package to avoid stale fixed includes. A solution that might be a little more clever would be to run fixincl.sh only when headers has been installed but I'm not sure if you know that in portage. Comments? I haven't looked very much at the portage source, and I really don't know python (yet) but I might have a look at it.
Is this still a problem with 5.0.9-r ?
closing with 5.0.9
