[U-Boot] gcc-4.6 warnings

[U-Boot] gcc-4.6 warnings

[Kumar Gala]

We get a bunch of warnings like this with gcc-4.6.x:

e1000.c:4334:3: warning: variable 'x' set but not used [-Wunused-but-set-variable]

Are we get with adding -Wunused-but-set-variable so they don't show up?

- k

[U-Boot] gcc-4.6 warnings

[Mike Frysinger]

On Wednesday, August 24, 2011 14:29:43 Kumar Gala wrote:
> We get a bunch of warnings like this with gcc-4.6.x:
> 
> e1000.c:4334:3: warning: variable 'x' set but not used
> [-Wunused-but-set-variable]
> 
> Are we get with adding -Wunused-but-set-variable so they don't show up?

in general, i like this new warning as it tends to catch variables that are 
dead.  i wonder why the e1000 driver has this indirection in the first place.  
simply calling readl() without using the return value shouldnt cause warnings 
or DCE ...
-mike
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 836 bytes
Desc: This is a digitally signed message part.
Url : http://lists.denx.de/pipermail/u-boot/attachments/20110824/730b29df/attachment.pgp 

[U-Boot] gcc-4.6 warnings

[Wolfgang Denk]

Dear Kumar Gala,

In message <3AA0E5B6-7E38-4CB0-94E2-F7BBA9A100FA@kernel.crashing.org> you wrote:
> We get a bunch of warnings like this with gcc-4.6.x:
> 
> e1000.c:4334:3: warning: variable 'x' set but not used [-Wunused-but-set-variable]
> 
> Are we get with adding -Wunused-but-set-variable so they don't show up?

If the variable is not used, why don't we remove it, then?

Best regards,

Wolfgang Denk

-- 
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
C makes it easy for you to shoot yourself in the foot. C++ makes that
harder, but when you do, it blows away your whole leg.
                                                 -- Bjarne Stroustrup

[U-Boot] gcc-4.6 warnings

[Kumar Gala]

On Aug 24, 2011, at 2:05 PM, Wolfgang Denk wrote:

> Dear Kumar Gala,
> 
> In message <3AA0E5B6-7E38-4CB0-94E2-F7BBA9A100FA@kernel.crashing.org> you wrote:
>> We get a bunch of warnings like this with gcc-4.6.x:
>> 
>> e1000.c:4334:3: warning: variable 'x' set but not used [-Wunused-but-set-variable]
>> 
>> Are we get with adding -Wunused-but-set-variable so they don't show up?
> 
> If the variable is not used, why don't we remove it, then?

In the vast number of cases it because of some #ifdef case not be defined in the given build.

- k

[U-Boot] gcc-4.6 warnings

[Wolfgang Denk]

Dear Kumar Gala,

In message <348935C0-0C2D-4A7A-8ABE-9D09E2904B26@kernel.crashing.org> you wrote:
> 
> > If the variable is not used, why don't we remove it, then?
> 
> In the vast number of cases it because of some #ifdef case not be
> defined in the given build.

well, that's the reason for the warnings showing up, it isn't the
reason why we cannot fix these?

Best regards,

Wolfgang Denk

-- 
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
"I used to think that the brain was the most wonderful  organ  in  my
body. Then I realized who was telling me this."        - Emo Phillips

[U-Boot] gcc-4.6 warnings

[Kumar Gala]

On Aug 25, 2011, at 12:52 AM, Wolfgang Denk wrote:

> Dear Kumar Gala,
> 
> In message <348935C0-0C2D-4A7A-8ABE-9D09E2904B26@kernel.crashing.org> you wrote:
>> 
>>> If the variable is not used, why don't we remove it, then?
>> 
>> In the vast number of cases it because of some #ifdef case not be
>> defined in the given build.
> 
> well, that's the reason for the warnings showing up, it isn't the
> reason why we cannot fix these?

:), Thus my query if we really wanted to try and fix them by adding more #ifdef's or live with that as an acceptable thing to do.

- k

[U-Boot] gcc-4.6 warnings

[Mike Frysinger]

On Thursday, August 25, 2011 10:27:42 Kumar Gala wrote:
> On Aug 25, 2011, at 12:52 AM, Wolfgang Denk wrote:
> > Kumar Gala wrote:
> >>> If the variable is not used, why don't we remove it, then?
> >> 
> >> In the vast number of cases it because of some #ifdef case not be
> >> defined in the given build.
> > 
> > well, that's the reason for the warnings showing up, it isn't the
> > reason why we cannot fix these?
> :
> :), Thus my query if we really wanted to try and fix them by adding more
> :#ifdef's or live with that as an acceptable thing to do.

i feel like some (many?) #ifdef's in the tree could be done without ifdefs (by 
relying on gcc's DCE) thus improving overall code quality
-mike
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 836 bytes
Desc: This is a digitally signed message part.
Url : http://lists.denx.de/pipermail/u-boot/attachments/20110825/6669ba99/attachment.pgp 

[U-Boot] gcc-4.6 warnings

[Tabi Timur-B04825]

On Thu, Aug 25, 2011 at 9:51 AM, Mike Frysinger <vapier@gentoo.org> wrote:

> i feel like some (many?) #ifdef's in the tree could be done without ifdefs (by
> relying on gcc's DCE) thus improving overall code quality

What's DCE?

-- 
Timur Tabi
Linux kernel developer at Freescale

[U-Boot] gcc-4.6 warnings

[Mike Frysinger]

On Thursday, August 25, 2011 15:50:57 Tabi Timur-B04825 wrote:
> On Thu, Aug 25, 2011 at 9:51 AM, Mike Frysinger <vapier@gentoo.org> wrote:
> > i feel like some (many?) #ifdef's in the tree could be done without
> > ifdefs (by relying on gcc's DCE) thus improving overall code quality
> 
> What's DCE?

dead code elimination
http://en.wikipedia.org/wiki/Dead_code_elimination

if we let the optimizer do it instead of the preprocessor, we get better code 
coverage in the face of different config settings.
-mike
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 836 bytes
Desc: This is a digitally signed message part.
Url : http://lists.denx.de/pipermail/u-boot/attachments/20110825/dfdcf6e1/attachment.pgp 

[U-Boot] gcc-4.6 warnings

[Wolfgang Denk]

Dear Kumar Gala,

In message <7056F5B1-AD6C-459E-80F1-8EE436CC781A@kernel.crashing.org> you wrote:
> 
> > well, that's the reason for the warnings showing up, it isn't the
> > reason why we cannot fix these?
>
> :), Thus my query if we really wanted to try and fix them by adding more
> #ifdef's or live with that as an acceptable thing to do.

My experience is that hushing up GCC warnings is not really a good
idea.  First, you will then also fail to see when new code adds such
problems.  Second, there have been cases where warnings in one GCC
version because serious bugs in later versions, so we better fix them.

Best regards,

Wolfgang Denk

-- 
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
It is a good thing for an uneducated man to read books of quotations.
                        - Sir Winston Churchill _My Early Life_ ch. 9

[U-Boot] gcc-4.6 warnings

[Timur Tabi]

Mike Frysinger wrote:
> if we let the optimizer do it instead of the preprocessor, we get better code 
> coverage in the face of different config settings.

Oh, I thought this was some new feature of U-Boot.

Can you give me an example of where DCE could be used to eliminate an #ifdef?

-- 
Timur Tabi
Linux kernel developer at Freescale

[U-Boot] gcc-4.6 warnings

[Timur Tabi]

Timur Tabi wrote:
> Oh, I thought this was some new feature of U-Boot.
> 
I meant a new feature of gcc. Ugh.

-- 
Timur Tabi
Linux kernel developer at Freescale

[U-Boot] gcc-4.6 warnings

[Mike Frysinger]

On Thursday, August 25, 2011 17:44:58 Timur Tabi wrote:
> Mike Frysinger wrote:
> > if we let the optimizer do it instead of the preprocessor, we get better
> > code coverage in the face of different config settings.
> 
> Oh, I thought this was some new feature of U-Boot.
> 
> Can you give me an example of where DCE could be used to eliminate an
> #ifdef?

a change i talked about but havent yet posted would be to fix debug().  
currently it looks more or less like:
#ifdef DEBUG
# define debug(...) printf(...)
#else
# define debug(...)
#endif
this causes people to do things like put local variables only used with 
debug() code behind an #ifdef DEBUG in their code otherwise they get unused 
variable warnings.

if, instead, we did something like:
#ifdef DEBUG
# define __DEBUG_KNOB 1
#else
# define __DEBUG_KNOB 0
#endif
#define debug(...) do { if (__DEBUG_KNOB) printf(...); } while (0)

the code would always get compiled and checked, but gcc would throw away it at 
DCE time, and there would be no unused variable warnings.
-mike
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 836 bytes
Desc: This is a digitally signed message part.
Url : http://lists.denx.de/pipermail/u-boot/attachments/20110825/31d98c01/attachment.pgp