Date: Mon, 01 Dec 2003 16:54:15 +0900 From: Alexander Nedotsukov <bland@FreeBSD.org> To: Jeremy Messenger <mezz7@cox.net> Cc: freebsd-gnome@FreeBSD.org Subject: Re: gnome-system-monitor (2.5.1) crashed on 5.2-BETA Message-ID: <3FCAF3A7.3050202@FreeBSD.org> In-Reply-To: <oprzhpbzc88ckrg5@smtp.central.cox.net> References: <oprzhl3ngb8ckrg5@smtp.central.cox.net> <3FCAD934.8050707@FreeBSD.org> <oprzhnjakq8ckrg5@smtp.central.cox.net> <3FCAE208.6020809@FreeBSD.org> <oprzhpbzc88ckrg5@smtp.central.cox.net>
next in thread | previous in thread | raw e-mail | index | archive | help
Jeremy Messenger wrote: > On Mon, 01 Dec 2003 15:39:04 +0900, Alexander Nedotsukov > <bland@FreeBSD.org> wrote: > >> Jeremy Messenger wrote: >> >>> On Mon, 01 Dec 2003 15:01:24 +0900, Alexander Nedotsukov >>> <bland@FreeBSD.org> wrote: >>> >>>> Jeremy Messenger wrote: >>>> >>>>> To reproduce is... just run it and it will crash. Here's backtraces: >>>>> >>>>> ================================= >>>> > <snip> > >>>>> ================================= >>>>> >>>>> Let me know if you think I should report to Gnome Bugzilla. >>>> >>>> >>>> >>>> Jeremy, >>>> >>>> From this backtrace I can only say that actual problem not in >>>> g-s-m itself but in libgtop2. Looks like at this point >>>> libgtop_server2 gcrashed for some reason. This SIGPIPE just an >>>> indication that communication pipe to it get lost. Of course it >>>> also have a sence to do better error handling in g-s-m. To see >>>> where it crashes please run cont from the gdb prompt till you get >>>> some serious signal like SIGABRT/BUS etc delivered. I think it will >>>> be some kind of assert(). >>> >>> >>> >>> I just ran 'run cont' and I still get the same result.. >>> >>> ================================= >>> (gdb) run cont >>> Starting program: /usr/X11R6/bin/gnome-system-monitor cont >>> LibGTop-Server: recv: Resource temporarily unavailable >>> >>> Program received signal SIGPIPE, Broken pipe. >>> 0x28c186cf in write () from /lib/libc.so.5 >>> >>> [...same backtraces as above...] >> >> >> Ah. Sorry. I mean you shoud first "run" g-s-m under gdb and get >> SIGPIPE then type in "cont" in the command line and press enter. "run >> cont" just runs g-s-m with parameter "cont" in the arg list. >> But I pretty sure now that this happens due libgtop simply aborts on >> any serious error on libtop_server2 communication pipe. This is by >> design. > > > Heh, I am still not skill on gdb, but I am trying.. :-) Here's info: > > ================================= > (gdb) cont > Continuing. > glibtop: write 40 bytes: Broken pipe > Error accessing memory address 0x284711a0: No such process. Yeah. Thats excactly the place. Let's focus on libgtop_server2 crash new. > > (gdb) x 0x284711a0 > 0x284711a0 <_thread_run>: Error accessing memory address > 0x284711a0: No such process. > ================================= > >>> ================================= >>> >>> So, I tried to run libgtop_server2 instead and I get this: >>> >>> ================================= >>> (gdb) run >>> Starting program: /usr/X11R6/bin/libgtop_server2 >>> kvm_open: /dev/mem: Permission denied >>> LibGTop-Server: kvm_open: Permission denied >> >> >> Thats ok since you runned suid binary under gdb as regular user. Try >> it as root. To nail down crash point in libtop_server2 I can advise >> to step through g-s-m till the moment libgtop initialized than run >> separate gdb copy and attach to libgtop_server2 process. Resume >> execution in both gdb sessions and see what happens next. >> Btw does some other libgtop2 client works on your machine? Let's say >> load monitor appet from gnomeappets or netspeed applet? > > > Nope, load monitor doesn't work, which it will crash.. I can't even > install gDesklets but might be due to libgtop's api/abi change. I am > not sure.. However, I just installed the netspeed applet and it can't > build; it might be same reason with gDesklets.. Bad. We really need to know what is going inside libgtop_server2. If you have a time to go througth the steps I described above please send me backtrace you get. Otherwise you have to wait till someone will do. I'll be available to look into this myself not early nearest weekend. All the best, Alexander.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?3FCAF3A7.3050202>