RE: [Nautilus-list] Re: oaf query sorting not working (was Strang e behaviour)

["Blad, John Erling" <john erling blad aftenposten no>]

Not sure but I've heard that gcc can change the order of evaluation
of the different function arguments. Somewhere in the docs it says
this is okey but I gess it's not so. ;/

Some of the sort algoritms use global storage and if you can't say
which argument is evaluated first the algorithm will fail.

This should be visible at high optimization level, don't really know
at which level it kick's in.. and sure it will be sensitive to
omit-frame-pointer and strenght-reduce.

I also had a weird bug once when the lt-operator was changed to an
ge-operator at some (O3?) optimization level. I had to write it as
!(a>=b) and then it worked. Later it seems like that bug went away.

John

> -----Original Message-----
> From: Frederic Crozat [mailto:fcrozat mandrakesoft com]
> Sent: Wednesday, August 08, 2001 10:59 AM
> To: Darin Adler
> Cc: gnome-vfs ximian com; nautilus-list lists eazel com;
> gnome-components-list gnome org
> Subject: [Nautilus-list] Re: oaf query sorting not working 
> (was Strange
> behaviour)
> 
> 
> On 17 Jul 2001 07:10:33 -0700, Darin Adler wrote:
> > On Tuesday, July 17, 2001, at 02:16  AM, Frederic Crozat wrote:
> > 
> > > I didn't knew about oaf-run-query but I've made some 
> tests with it :
> > >
> > > running the same query with both order give me :
> > >
> > > Number of sort criteria: 1
> > > prefer_by_list_order(iid,
> > > 
> ['OAFIID:nautilus_mozilla_content_view:1ee70717-57bf-4079-aae5
> -922abdd576b1'
> > > 
> ,'OAFIID:nautilus_text_view:fa466311-17c1-435c-8231-c9fc434b6437'])
> > > number of results: 2
> > > OAFIID:nautilus_text_view:fa466311-17c1-435c-8231-c9fc434b6437
> > > 
> OAFIID:nautilus_mozilla_content_view:1ee70717-57bf-4079-aae5-9
> 22abdd576b1
> > >
> > > Number of sort criteria: 1
> > > prefer_by_list_order(iid,
> > > 
> ['OAFIID:nautilus_text_view:fa466311-17c1-435c-8231-c9fc434b64
> 37','OAFIID:
> > > 
> nautilus_mozilla_content_view:1ee70717-57bf-4079-aae5-922abdd576b1'])
> > > number of results: 2
> > > OAFIID:nautilus_text_view:fa466311-17c1-435c-8231-c9fc434b6437
> > > 
> OAFIID:nautilus_mozilla_content_view:1ee70717-57bf-4079-aae5-9
> 22abdd576b1
> > >
> > > I have this problem with both oaf 0.6.5 and a oaf 
> snapshot from 20010711
> > > (oaf-stable-0-6 branch, needed to correct activation problem with
> > > evolution..)
> > >
> > > Any idea ?
> > 
> > If you're comfortable running under gdb you can try putting 
> the oafd 
> > process under the debugger while you run the query with 
> oaf-run-query. I'd 
> > set a breakpoint on qexp_func_prefer_by_list_order and see 
> what happens in 
> > there. Or you could add printfs to that function and then 
> kill and restart 
> > oafd and debug that way.
> > 
> > [If not, maybe you can set up a way for me to ssh to that 
> machine so I can 
> > debug it, or perhaps we can have a debugging session with 
> us talking via 
> > IRC, but that probably won't be today.]
> 
> Ok, I've finally found were this strange bug come from !!
> 
> This strange bug appears only when we compile oaf with gcc 
> 2.96 and "-O3
> -fomit-frame-pointer -pipe -mcpu=pentiumpro -march=i586 -ffast-math
> -fno-strength-reduce" :((
> 
> This bug doesn't appears with egcs nor gcc 3.0.. I've given 
> this problem
> to our Mdk gcc specialist :))
> 
> 
> 
> -- 
> Frédéric Crozat
> MandrakeSoft
> 
> 
> _______________________________________________
> Nautilus-list mailing list
> Nautilus-list lists eazel com
> http://lists.eazel.com/mailman/listinfo/nautilus-list
>