[Pdns-users] Automatic AXFR as slave with 2.9.10 doesn't seem to work right?

[Derrik Pates]

On Fri, Aug 22, 2003 at 10:44:37PM +0200, bert hubert wrote:
> This breaks the bonafide use of those fds by the guardian pdns_control
> socket proxy, which should then have to move to another fd entirely, like,
> 3.

Well, considering those fd's are always magic, and really should never
be used for anything else, shouldn't that be done anyway? And whatever
socket number you get is assigned when you use the socket() syscall, so
wouldn't hardcoding it, or making blind assumptions about what it is, be
bad from the start?

> Your patch is fine as far as it goes but it only papers over the problem -
> I'd in fact prefer pdns to die if it tries to write to stdout.

Well, it didn't die before, just polluted other streams, causing a
problem that I wouldn't have ever noticed if I wasn't watching the logs
carefully and noticing that "oh, wait, it never axfr's those domains
until I force one. something's not right."

> Currently, we at least experience problems when writing to stdout, helping
> us to notice the problem.

As I said, an obscure one that might have gone unnoticed otherwise. I'd
rather it behave correctly and spit stuff into /dev/null, rather than
cause subtle problems like that.

-- 
Derrik Pates
dpates at dsdk12.net
dpates at voxel.net