INN 18-10 coredump

[Miquel van Smoorenburg]

We have been seeing intermittent coredumps of innd. Well, no real
cores as ulimit -c was set to zero. Today I finally got a real core,
and this is what it shows:


CVS version of 18-10:

Core was generated by `/usr/local/news/bin/innd -p4'.
Program terminated with signal 11, Segmentation fault.
#0  SITEwantsgroup (sp=0x9336888, name=0x812e200 "nctu.test") at site.c:915
915         for (argv = sp->Patterns; (pat = *argv++) != NULL; ) {
(gdb) bt
#0  SITEwantsgroup (sp=0x9336888, name=0x812e200 "nctu.test") at site.c:915
#1  0x805e012 in ARTsendthegroup (name=0x812e200 "nctu.test") at art.c:1292
#2  0x805fa3b in ARTpost (cp=0x401b7468) at art.c:2239
#3  0x80691c9 in NCpostit (cp=0x401b7468) at nc.c:196
#4  0x806a530 in NCproc (cp=0x401b7468) at nc.c:927
#5  0x806ac8c in NCreader (cp=0x401b7468) at nc.c:1154
#6  0x806569c in CHANreadloop () at chan.c:973
#7  0x8068b89 in main (ac=2, av=0xbffffe74) at innd.c:940
(gdb) print sp->Patterns
$2 = (char **) 0x0
(gdb) up
#1  0x805e012 in ARTsendthegroup (name=0x812e200 "nctu.test") at art.c:1292
1292            if (sp->Name != NULL && SITEwantsgroup(sp, name)) {
(gdb) print sp->Name
$3 = 0x92decc0 "1``(\"`@$#`@,%!@4\"!p$````!`a$#(3$$$d$%41,b8084,g&q(x&1\r\nmh='p%19\"4l$sx4-b<x*#dm+q_]h`# ,!``(1`q$`/p#@>n=/@z7=p16k2r13\r\nm65o=`s8u#q(py&vvv:]3u;i<6#m>./;n:=-o\\,j\\md8y-\\f4hu.,g->7##dr\r\nm1<h1;2y^ahvd:me;b5"...
(gdb) l
1287        SITE                *sp;
1288        int                 i;
1289        NEWSGROUP           *ngp;
1290
1291        for (ngp = NGfind(ARTctl), sp = Sites, i = nSites; --i >= 0; sp++)
1292            if (sp->Name != NULL && SITEwantsgroup(sp, name)) {
1293                SITEmark(sp, ngp);
1294            }
1295    }
1296
(gdb) print Sites[0].Name
$9 = 0x8a4f6b0 "controlchan!"
(gdb) print Sites[1].Name
$10 = 0x97a8788 "innfeed!"
(gdb) print Sites[2].Name
$11 = 0x96372a0 "innfeed1!"
(gdb) print Sites[3].Name
$12 = 0x94621b0 "innfeed2!"
(gdb) print Sites[4].Name
$13 = 0x959f420 "innfeed3!"
(gdb) print Sites[5].Name
$14 = 0x9666440 "innfeed4!"
(gdb) print Sites[6].Name
$15 = 0x92decc0 "1``(\"`@$#`@,%!@4\"!p$````!`a$#(3$$$d$%41,b8084,g&q(x&1\r\nmh='p%19\"4l$sx4-b<x*#dm+q_]h`# ,!``(1`q$`/p#@>n=/@z7=p16k2r13\r\nm65o=`s8u#q(py&vvv:]3u;i<6#m>./;n:=-o\\,j\\md8y-\\f4hu.,g->7##dr\r\nm1<h1;2y^ahvd:me;b5"...
(gdb) print Sites[7].Name
$16 = 0x90cae48 "news2.cistron.nl"
(gdb) print Sites[8].Name
$17 = 0x8166ff0 "newsfeed.isdnet.net"

.. looks like Sites[6] got corrupted. It should have been "news.cistron.nl".
Weird ?

Any idea what could cause this?

Mike.
-- 
First things first, but not necessarily in that order.
-- 
The From: and Reply-To: addresses are internal news2mail gateway addresses.
Reply to the list or to miquels at cistron.nl (Miquel van Smoorenburg)