Re: pahole displays type incorrectly

Re: pahole displays type incorrectly

[Arnaldo Carvalho de Melo]

Em Mon, Aug 17, 2009 at 12:42:28PM -0300, Arnaldo Carvalho de Melo escreveu:
> Em Mon, Aug 17, 2009 at 04:12:25PM +0200, Jan Engelhardt escreveu:
> > Hi,
> > 
> > 
> > given the following struct:
> > 
> >   struct x {
> >   	const char *const s;
> >   };
> > 
> > pahole will display it as:
> > 
> >   struct x {
> >         const const char *s; /* .. */
> >   };
> > 
> > And that is not quite right. The sparse(1) tool would say "how much more 
> > const do you want it to be?" :)
> > I used pahole.git commit f0435286c0b.
> 
> pahole is not properly handling that case at the moment, the DWARF chain
> is:
> 
> [acme@doppio pahole]$ readelf -wi object_samples/jengelh\@medozas.de/const_const
> <SNIP>
>  <1><2d>: Abbrev Number: 2 (DW_TAG_structure_type)
>     <2e>   DW_AT_name        : x        
>     <30>   DW_AT_byte_size   : 8        
>     <31>   DW_AT_decl_file   : 1        
>     <32>   DW_AT_decl_line   : 1        
>     <33>   DW_AT_sibling     : <0x44>   
>  <2><37>: Abbrev Number: 3 (DW_TAG_member)
>     <38>   DW_AT_name        : s        
>     <3a>   DW_AT_decl_file   : 1        
>     <3b>   DW_AT_decl_line   : 2        
>     <3c>   DW_AT_type        : <0x44>   
>     <40>   DW_AT_data_member_location: 2 byte block: 23 0
> (DW_OP_plus_uconst: 0)
>  <1><44>: Abbrev Number: 4 (DW_TAG_const_type)
>     <45>   DW_AT_type        : <0x49>   
>  <1><49>: Abbrev Number: 5 (DW_TAG_pointer_type)
>     <4a>   DW_AT_byte_size   : 8        
>     <4b>   DW_AT_type        : <0x4f>   
>  <1><4f>: Abbrev Number: 4 (DW_TAG_const_type)
>     <50>   DW_AT_type        : <0x54>   
>  <1><54>: Abbrev Number: 6 (DW_TAG_base_type)
>     <55>   DW_AT_byte_size   : 1        
>     <56>   DW_AT_encoding    : 6        (signed char)
>     <57>   DW_AT_name        : (indirect string, offset: 0x72): char    
> 
> <SNIP>
> 
> So if we find a const at the start of the chain, we must render it
> first, then the type, but if we find it inside the chain, it has to be
> rendered at the end of the chain... I'm reworking the code that
> recursively traverses the chain to fix this, but I'm being sidetracked
> with some other work, stay tuned and I'll get this fixed today
> 
> Thanks for the report!

Sorry for the delay, should be fixed by the latest commit, please give
it a try!

http://git.kernel.org/?p=linux/kernel/git/acme/pahole.git;a=commitdiff;h=846927a3f87f730e0ae201b5f0b3b46212a7bb8e

- Arnaldo
--
To unsubscribe from this list: send the line "unsubscribe dwarves" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: pahole displays type incorrectly

[Arnaldo Carvalho de Melo]

Em Thu, Sep 10, 2009 at 07:22:56PM -0300, Arnaldo Carvalho de Melo escreveu:
> Em Thu, Sep 10, 2009 at 05:35:17PM -0300, Arnaldo Carvalho de Melo escreveu:
> > > So if we find a const at the start of the chain, we must render it
> > > first, then the type, but if we find it inside the chain, it has to be
> > > rendered at the end of the chain... I'm reworking the code that
> > > recursively traverses the chain to fix this, but I'm being sidetracked
> > > with some other work, stay tuned and I'll get this fixed today
> > > 
> > > Thanks for the report!
> > 
> > Sorry for the delay, should be fixed by the latest commit, please give
> > it a try!
> > 
> > http://git.kernel.org/?p=linux/kernel/git/acme/pahole.git;a=commitdiff;h=846927a3f87f730e0ae201b5f0b3b46212a7bb8e
> 
> Bogus patch, right fix should come soon.

Please do a git pull and see if it now works properly.

Thanks,

- Arnaldo
--
To unsubscribe from this list: send the line "unsubscribe dwarves" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: pahole displays type incorrectly

[Jan Engelhardt]

On Thursday 2009-09-10 22:35, Arnaldo Carvalho de Melo wrote:
>
>Sorry for the delay, should be fixed by the latest commit, please give
>it a try!
>
>http://git.kernel.org/?p=linux/kernel/git/acme/pahole.git;a=commitdiff;h=846927a3f87f730e0ae201b5f0b3b46212a7bb8e

Yes it does work. The testcase seems to be reducable to just the
struct without a function, but still with a non-static definition
(instance) of the struct. I.e. on the file level:

struct foo {
	const char *const z;
};
struct foo x;
--
To unsubscribe from this list: send the line "unsubscribe dwarves" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: pahole displays type incorrectly

[Arnaldo Carvalho de Melo]

Em Sat, Sep 12, 2009 at 03:41:56AM +0200, Jan Engelhardt escreveu:
> 
> On Thursday 2009-09-10 22:35, Arnaldo Carvalho de Melo wrote:
> >
> >Sorry for the delay, should be fixed by the latest commit, please give
> >it a try!
> >
> >http://git.kernel.org/?p=linux/kernel/git/acme/pahole.git;a=commitdiff;h=846927a3f87f730e0ae201b5f0b3b46212a7bb8e
> 
> Yes it does work. The testcase seems to be reducable to just the
> struct without a function, but still with a non-static definition
> (instance) of the struct. I.e. on the file level:
> 
> struct foo {
> 	const char *const z;
> };
> struct foo x;

Thanks for testing! I'll update the test case to reduce it.

- Arnaldo
--
To unsubscribe from this list: send the line "unsubscribe dwarves" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: pahole displays type incorrectly

[Arnaldo Carvalho de Melo]

Em Sat, Sep 12, 2009 at 10:24:05AM -0300, Arnaldo Carvalho de Melo escreveu:
> Em Sat, Sep 12, 2009 at 03:41:56AM +0200, Jan Engelhardt escreveu:
> > 
> > On Thursday 2009-09-10 22:35, Arnaldo Carvalho de Melo wrote:
> > >
> > >Sorry for the delay, should be fixed by the latest commit, please give
> > >it a try!
> > >
> > >http://git.kernel.org/?p=linux/kernel/git/acme/pahole.git;a=commitdiff;h=846927a3f87f730e0ae201b5f0b3b46212a7bb8e
> > 
> > Yes it does work. The testcase seems to be reducable to just the
> > struct without a function, but still with a non-static definition
> > (instance) of the struct. I.e. on the file level:
> > 
> > struct foo {
> > 	const char *const z;
> > };
> > struct foo x;
> 
> Thanks for testing! I'll update the test case to reduce it.

While running regtest I noticed it is still buggy :-\ It works when you
have a "const char *const x"  case, but reverses both "const char *x"
and "char *const x", but now parents are visiting, time to go AFK :-)

If anybody feels like looking at the latest commits to see how it
should't be done and comes with a patch, I'd appreciate 8-)

- Arnaldo
--
To unsubscribe from this list: send the line "unsubscribe dwarves" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: pahole displays type incorrectly

[Arnaldo Carvalho de Melo]

Em Sat, Sep 12, 2009 at 11:49:39AM -0300, Arnaldo Carvalho de Melo escreveu:
> Em Sat, Sep 12, 2009 at 10:24:05AM -0300, Arnaldo Carvalho de Melo escreveu:
> > Em Sat, Sep 12, 2009 at 03:41:56AM +0200, Jan Engelhardt escreveu:
> > > 
> > > On Thursday 2009-09-10 22:35, Arnaldo Carvalho de Melo wrote:
> > > >
> > > >Sorry for the delay, should be fixed by the latest commit, please give
> > > >it a try!
> > > >
> > > >http://git.kernel.org/?p=linux/kernel/git/acme/pahole.git;a=commitdiff;h=846927a3f87f730e0ae201b5f0b3b46212a7bb8e
> > > 
> > > Yes it does work. The testcase seems to be reducable to just the
> > > struct without a function, but still with a non-static definition
> > > (instance) of the struct. I.e. on the file level:
> > > 
> > > struct foo {
> > > 	const char *const z;
> > > };
> > > struct foo x;
> > 
> > Thanks for testing! I'll update the test case to reduce it.
> 
> While running regtest I noticed it is still buggy :-\ It works when you
> have a "const char *const x"  case, but reverses both "const char *x"
> and "char *const x", but now parents are visiting, time to go AFK :-)
> 
> If anybody feels like looking at the latest commits to see how it
> should't be done and comes with a patch, I'd appreciate 8-)

Should be ok now. See lest cset, with test cases for:

const char *x;
char *const x;
const char *const x;

- Arnaldo
--
To unsubscribe from this list: send the line "unsubscribe dwarves" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html