From: Brian D H. <bdh...@c4...> - 2001-07-03 17:18:22
|
Hamish, I'm testing on a machine without a TNC so I would have missed that. If you patch it in do you see your beacons flowing properly? I spent last night putting triggers around the Add/DeleteSession code and watching the counters, loop index, and various lock behaviors. I've not found anything yet that doesn't appear correct. I'm working on a theory that on a box with a very high connect/disconnect rate (such as first), you get into a race where the DeleteSession code hasn't yet freed a session (socket set to "-1") when AddSession comes looking for a new one. I'm able to occasionally recreate the deadlock by setting MaxUsers down to 60 (17 IGATES) and doing the following repeatedly: start 20 Xastir sessions, "killall -9 xastir", start over again. If I do it fast enough I can deadlock. The deadlock isn't where I expect it though. (of course!) I think more navel contemplation is in order today. -- ============================================================ Brian D Heaton | I fear that we have awakened Principal Consultant | a sleeping giant and instilled C4I2.com System Consultants | in him a terrible resolve. bdh...@c4... | -- Admiral Isoruku Yamamoto USA (719) 623-0381 | -- Imperial Japanese Navy UK +44 (0)845 127-5400 | -- December 7, 1941 On 2001.07.03 08:55 Hamish Moffatt wrote: > On Wed, Jul 04, 2001 at 12:46:25AM +1000, Hamish Moffatt wrote: > > It handles the beacon strings differently in some way than > > the previous release did, and the end result is that > > TncBeacon->data is not valid. > > Me again. The aprsString constructorSetUp() code has the following: > > if (path.find(">") == npos) { > aprsType = APRSERROR; // then it's bogus > return; > } > > This disqualifies plain old first-party packets which might be used > for the TncBeacon. I think it should be changed to > > if ((sourceSock != SRC_INTERNAL) && (path.find(">") == npos)) > { > aprsType = APRSERROR; // then it's bogus > return; > } > > However, I don't know this area of the code very well so I ask for > your input (Chuck and Brian) first. > > > thanks, > Hamish > -- > Hamish Moffatt VK3SB <ha...@de...> <ha...@cl...> > > _______________________________________________ > Aprsd-devel mailing list > Apr...@li... > http://lists.sourceforge.net/lists/listinfo/aprsd-devel > |