Segmentation fault in FreeBSD 9.0 flockfile ()

[Unga]

> From: Robert Bonomi <bonomi at mail.r-bonomi.com>
> To: freebsd-questions at freebsd.org
> Cc: 
> Sent: Wednesday, May 2, 2012 6:38 AM
> Subject: Re: Segmentation fault in FreeBSD 9.0 flockfile ()
> 
> 
> Unga <unga888 at yahoo.com> wrote;
>>  From: Darren Baginski <kickbsd at yandex.com>
>>  > Unga <unga888 at yahoo.com> wrote:
>>  > 
>>  >01.05.2012, 22:08, "Unga" <unga888 at yahoo.com>:
>>  >> Hi all
>>  >>
>>  >> I'm getting a  Segmentation fault in FreeBSD 9.0 as follows 
> for myprog.c:
>>  >>
>>  >> Reading symbols from /libexec/ld-elf.so.1...done.
>>  >> Loaded symbols for /libexec/ld-elf.so.1
>>  >> #0  0x28ebb062 in flockfile () from /lib/libc.so.7
>>  >> [New Thread 29c04900 (LWP 100575/SDLTimer)]
>>  >> [New Thread 29c04300 (LWP 100416/myprog)]
>>  >> (gdb)
>>  >> (gdb) info threads
>>  >> * 2 Thread 29c04300 (LWP 100416/myprog)  0x28ebb062 in flockfile 
> ()
>>  >>    from /lib/libc.so.7
>>  >>   1 Thread 29c04900 (LWP 100575/SDLTimer)  0x28e1527b in _umtx_op 
> ()
>>  >>    from /lib/libc.so.7
>>  >> (gdb)
>>  >>
>>  >> I don't use flockfile () directly in my program.
>>  >>
>>  >
>>  >Hi!
>>  >
>>  >Mind to share code snippet caused the problem?
>> 
>>  Hi Thanks for the reply. I have mentioned in my original post that neithe
>>  r I nor SDL use the flockfile () in the source.
>> 
> 
> There was a _reason_ the prior poster asked about seeing your code.
> The information you provided is not adequate/sufficient for anyone to
> have any chance of helping you.  Informationn that -is- necessary was
> therefore requested.  
> 
> What you "don't use" is irrelevant to solving the problem. 
> 
> To be any help one must figure out what you _did_ do -- which =indirectly= 
> calls flockfile().
> 
> Your code calls "something" in 'libc'  that calls flockfile().
> 
> You "don't know" what that something in your code is.  Other eyes 
> _might_
> recognize it, -if- they saw what you were doing.
> 
> Here is how to find out what does that calling, yourself.
> 1) recompile all the components of 'myprog' with the '-g' 
> compiler option.
> 2) start 'script', so that you have a complete log of the debugger 
> session.
> 3) load the debugger with 'myprog' as the program to be debugged.
> 4) set a breakpoinnt in the 'flockfile' function
> 5) 'run' the program
> 6) when it stops at flockfile() entry. use 'where' to see a stack trace.
> 7) 'display' each of the arguments to flockfile().
> 9) go 'up' to the calling code.
> 9) repeat steps 7 and 8 util you reach 'your' code.
> 10) now 'continue' the program.  If it immediately segfaults, this was 
> the
>     offending call.  Otherwise, the program will stop aggain at another
>     flockfile() invocation. repeat from step 7 -until- you get the segfault.
> 
> If, from that debugger session, you still don't see anything wrong, post
> the log of the debugger session.
> 
> There is a *remote* possibility that the program will run properly under
> the debugger, while it crashes when run directly.  This has been known 
> to happen when a program _uses_ a variable before initializing it.
> 
> *IF* that happens, you will have to modify 'myprog' to add diagnostic
> output before and after each identified point (above) in your code that
> indirectly invokes flockfile() to find which is failing.  
> 
> In _that_ event, you =will= have to post the 'relevant' source-code from
> 'myprog', with the failing statement identified.
> 
Hopefully I have not offended anyone with my reply :) Very sorry if I did.

My this FreeBSD 9.0 box is seems very messy. I have another issue that getaddrinfo(3) doesn't accept "localhost". 

After installing FreeBSD 9.0, I installed KDE, then realised that the existing Intel display driver does not support my Intel Sandy Bridge based GPU. So I had to patch the kernel to support KMS based latest Intel driver, upgraded to Xorg to 7.5.2, etc. etc. Whole process was done multiple times as I was new to this complex upgrade.

I presume, I should erase off the hard disk and reinstall everything cleanly and then retry my programs on FreeBSD 9.0. If the problem still persists, I'll surely do a detailed debug. Pls note, whole thing works well on FreeBSD 8.1.

As this is time consuming and as per existing commitments, I only try this at least in another two weeks time. I will sure let the list know the outcome.

Best regards
Unga