Re: [Gzz] Tuukkah's generator problem

[Tuomas Lukka]

On Mon, Apr 07, 2003 at 12:30:51PM +0300, Tuukka Hastrup wrote:
> On Mon, 7 Apr 2003, Tuomas Lukka wrote:
> > [23:05:46] <tuukkah> libvob/src/jni$ ./Generator
> > [23:05:46] <tuukkah>         IrregularQuad
> > [23:05:46] <tuukkah> Virheellinen k?sky
> > [23:06:20] <tuukkah> "Virheellinen käsky" = Illegal instruction
> > [23:09:00] <tuukkah> (although "virheellinen" = flawed, and "käsky" = order)
> > [23:12:02] <tuukkah> and "make clean" doesn't even remove it
> > [23:13:57] <tuukkah> it was compiled into same form again :-(
> 
> I extended "make clean" so it removes Generator now, plus some other 
> generated files there was left.

Thank you.

> > Umm, it has to be run from libvob/, not libvob/src/jni, as you can
> > see from the Makefile. 
> 
> The point was that there should never come "Illegal instruction", 

Agreed.

> and as I 
> had said before, it seems I get the same signal even when it's run 
> properly from the Makefile.

Ok, *that*'s worrying!

> > And since there's debugging info you *could* of course have run gdb and
> > gotten us a stack trace...
> 
> Now that you suggested it, yes:
> 
> =0 address@hidden:~/cvs/Fenfire/libvob$ gdb src/jni/Generator
> GNU gdb 5.3-debian
> Copyright 2002 Free Software Foundation, Inc.
> GDB is free software, covered by the GNU General Public License, and you are
> welcome to change it and/or distribute copies of it under certain conditions.
> Type "show copying" to see the conditions.
> There is absolutely no warranty for GDB.  Type "show warranty" for details.
> This GDB was configured as "i386-linux"...
> (gdb) run
> Starting program: /home/tuukka/cvs/Fenfire/libvob/src/jni/Generator
> [New Thread 16384 (LWP 1970)]
>         IrregularQuad
>  
> Program received signal SIGILL, Illegal instruction.
> [Switching to Thread 16384 (LWP 1970)]
> __static_initialization_and_destruction_0 (__initialize_p=-1073743768,
>     __priority=-1073743728) at ../../include/vob/vobs/Irregu.hxx:46
> 46              freq = 1.0 / ripple_period;
> (gdb) bt
> #0  __static_initialization_and_destruction_0 (__initialize_p=-1073743768,
>     __priority=-1073743728) at ../../include/vob/vobs/Irregu.hxx:46
> #1  0x080705fb in 
> _GLOBAL__I__ZNSs15_M_replace_safeIPKcEERSsN9__gnu_cxx17__normal_iteratorIPcSsEES6_T_S7_Generator.cxxl87bjb
>  ()
>     at /usr/include/c++/3.2/bits/locale_facets.h:111
> #2  0x08058f4d in _init ()
> #3  0x40613a1f in __libc_start_main () from /lib/libc.so.6
> (gdb)
> 
> Does that say anything to you?

Waaaait, what CPU are you running on? 

Could it be that the architecture in rules.mk is wrong for your machine (P4, 
SSE2)?
Could you twiddle around there and see if it helps.

> > [23:22:37] <tuukkah> ridiculous idea: stop coding for one week, just write 
> > documentation
> > 
> > What documentation would have helped the current problem? If you tell me
> > what you'd like to read, I'll write it.
> 
> If "the current problem" is taken as my tries to compile everything, we 
> could start with a README in every subproject, stating at least the 
> obvious dependencies.

Ok, whose job should we make this be?

> > There's no point in saying just "write some documentation"; documentation
> > written from that starting point will have no focus and no flow.
> 
> The point was that at this point, any documentation would be beneficial 
> :-) 

Umm... I *REALLY* don't think you mean what you say there ;)

> There are common ideas about the minimum level of documentation, which 
> we certainly don't match yet. 

Common ideas -- why not PEG in fenfire or navidoc your idea of minimum
documentation and we'll get started from there.

> Further, I don't believe in someone telling what everybody should do, 

Well, anarchy hasn't worked too well for us either ;)

> like "If you tell me what you'd like to read, I'll write it." How should I 
> be able to enumerate everything?

Not everything but the things you need. That's the point of documentation.
Not saying EVERYTHING that can be truthfully said about the system in the common
logical framework, but saying what is needed for normal people to do normal 
tasks.

Documentation isn't something where you can say to a programmer
"write some documentation" and shut him in a room for 6 months ;)
What you get in the end is certainly not to the point, and most likely
relatively useless. Documents written in a vacuum are not fun to write
and even less fun to read.

It's more like asking me "I don't understand how the X build process works,
since after XXX it does YYY, could you document that?". *THEN* I can do 
something
that really helps you.

        Tuomas