Duarte Fonseca <[email protected]> wrote: > > 2009/11/12 Alexander Clouter <[email protected]>: > >> You should also compile the whole thing with optimisations turned off >> and debugging symbols in there; you are not doing the former so it might >> make it more difficult to work out what is wrong: >> ---- >> CFLAGS='-O0 -g' ./configure >> ---- > > Here is the gdb output and valgrind with these flags set. > > http://dl.dropbox.com/u/797051/radiusd/gdb-radiusd-2.log > http://dl.dropbox.com/u/797051/radiusd/valgrind-radiusd-2.log > This is a different implosion this time round to your last one. Looks like Alan's SQL patch simply postponed the SIGSEGV to a different part of the code. Not sure what should be done, but it looks like errorcode also needs to be changed (probably set to NULL explicitly or something). Alternatively it might be that the error code table in sql_postgresql.h needs updating? No real idea here, Alan?
I only was able to trigger this bug once, and that was during a spanning-tree going crazy (scheduled test of a reboot of one of our core routers) when FreeRADIUS would have been unable to talk to the SQL server. Although FreeRADIUS should not be dying, it does make me wonder why you can reproduce this problem so regularly :) Cheers -- Alexander Clouter .sigmonster says: Feminists just want the human race to be a tie. - List info/subscribe/unsubscribe? See http://www.freeradius.org/list/users.html
