On Sat, 2005-07-30 at 14:22 +0200, e09f6a7593f8ae3994ea57e1117f67ec wrote:
Balazs Scheidler wrote:
Thanks for the detailed problem report. I have added a different solution though, trying to merge the diverging code paths for SOCK_DGRAM and SOCK_STREAM. I've committed a fix to my tla archive, the snapshot for tomorrow should contain the fix, or you can apply the patch below to your tree directly.
Can you please test if this one works (I've compile tested and slightly runtime tested and it seems to work fine).
I have compiled syslog-ng-1.9.5+20050730.tar.gz and it does not seems to work and it cannot open udp source.
Debugging with a single udp() source:
afsocket_sd_init (s=0x8067f38, cfg=0x8062a20, persist=0x0) at afsocket.c:403 self->connections = persist_config_fetch(persist, afsocket_sd_format_persist_name(self, FALSE)); self->connections is null due persist==0x0
afsocket.c:412 if (!(self->flags & AFSOCKET_KEEP_ALIVE) || !self->connections || !afsocket_open_s... bails out here self->flags = 0x701, self->connection=null self->flags & AFSOCKET_KEEP_ALIVE > 0, hmmm, udp keepalive?
Note that persist if set to NULL at main.c:372 by calling cfg_init(cfg, NULL)
Sorry, did a stupid mistake, I tested a previous copy of the syslog-ng binary and not the one which already contained the changes. However I've extended the preliminary functional test program to also cover tcp/udp message sending and added some more fixes. Tomorrow's snapshot should contain the changes. PS: let me know if a public Arch archive would be useful in addition to the daily snapshots. -- Bazsi