patch for nnrpd-slaves mounting spool ro via NFS (tradindexd)

[bill davidsen]

In article <20010328232403.C1660 at serpens.de>,
S.P.Zeidler <spz at serpens.de> wrote:

| ---------------------- patch snip ------------------------- 
| *** ov3.c.orig	Sat Mar 24 01:17:22 2001
| --- ov3.c	Tue Mar 27 14:09:31 2001
	[... snip ...]
| ***************
| *** 179,187 ****
| --- 186,203 ----
|       groupfn = NEW(char, strlen(dirname) + strlen("/group.index") + 1);
|       strcpy(groupfn, dirname);
|       strcat(groupfn, "/group.index");
| + #ifdef SYNCGINDEXMMAP
| +     strncpy(GROUPfn, groupfn, PATH_MAX);
| + #endif
| + #ifndef NFSREADER
|       GROUPfd = open(groupfn, O_RDWR | O_CREAT, ARTFILE_MODE);
|       if (GROUPfd < 0) {
|   	syslog(L_FATAL, "tradindexed: could not create %s: %m", groupfn);
| + #else
| +     GROUPfd = open(groupfn, O_RDONLY, ARTFILE_MODE);
| +     if (GROUPfd < 0) {
| + 	syslog(L_FATAL, "tradindexed: could not open %s: %m", groupfn);
| + #endif
|   	DISPOSE(groupfn);
|   	return FALSE;
|       }

Okay, dumb question... is there something I missed in configure to set
this? I'm trying a compile with the flag set manually as we speak, but I
suspect this is a stopgap way to fo it.

More important, this should be a runtime option, not a compile time
option, since I want to mount the whole newsbase over NFS and just have
the slaves run with the flag set at runtime.

Since it will work (or not) as-is, no rush, but before I do this, is
there a good reason not to? I actually think this might justify an
option on the nnrpd command line, but that's hard to pass back to ov3.

Yes, I thought of the obvious to make group.index a symlink to
something really local... I used to run NFS slaves before, but the
performance wasn't good. Now that I have faster networks I want to look
at it again.

-- 
bill davidsen <davidsen at tmr.com>
  CTO, TMR Associates, Inc
Doing interesting things with little computers since 1979.