endian.h warnings

endian.h warnings

[Andre Renaud]

Hi,
When using musl with -Wparentheses enabled, I get the following warnings:
../include/endian.h: In function '__bswap32':
../include/endian.h:32:2: warning: suggest parentheses around
arithmetic in operand of '|' [-Wparentheses]
../include/endian.h:32:2: warning: suggest parentheses around
arithmetic in operand of '|' [-Wparentheses]
../include/endian.h: In function '__bswap64':
../include/endian.h:37:2: warning: suggest parentheses around '+'
inside '<<' [-Wparentheses]

I appreciate that these aren't bugs, and are just gcc being over
enthusiastic in it's warnings (which I've asked it to be on the
command line), however would it be worth sticking some parentheses
into the code simply to avoid this kind of noise being visible for
users?

Regards,
Andre

Re: endian.h warnings

[Rich Felker]

On Wed, Jul 10, 2013 at 02:21:58PM +1200, Andre Renaud wrote:
> Hi,
> When using musl with -Wparentheses enabled, I get the following warnings:
> .../include/endian.h: In function '__bswap32':
> .../include/endian.h:32:2: warning: suggest parentheses around
> arithmetic in operand of '|' [-Wparentheses]
> .../include/endian.h:32:2: warning: suggest parentheses around
> arithmetic in operand of '|' [-Wparentheses]
> .../include/endian.h: In function '__bswap64':
> .../include/endian.h:37:2: warning: suggest parentheses around '+'
> inside '<<' [-Wparentheses]
> 
> I appreciate that these aren't bugs, and are just gcc being over
> enthusiastic in it's warnings (which I've asked it to be on the
> command line), however would it be worth sticking some parentheses
> into the code simply to avoid this kind of noise being visible for
> users?

I really don't understand why gcc is generating warnings for these; I
thought it automatically suppressed warnings for system include
directories, and even macro expansion text generated from macros
defined in these headers. If that suppression is not working
correctly, maybe we should work around these warnings. I agree it's
problematic for users to get spammed with irrelevant warnings that
decrease the signal to noise ratio of the compiler's warning
generation.

Rich

Re: endian.h warnings

[Nathan McSween]

On Tuesday, July 09, 2013 7:21:58 PM, Andre Renaud wrote:
> Hi,
> When using musl with -Wparentheses enabled, I get the following warnings:
> ../include/endian.h: In function '__bswap32':
> ../include/endian.h:32:2: warning: suggest parentheses around
> arithmetic in operand of '|' [-Wparentheses]
> ../include/endian.h:32:2: warning: suggest parentheses around
> arithmetic in operand of '|' [-Wparentheses]
> ../include/endian.h: In function '__bswap64':
> ../include/endian.h:37:2: warning: suggest parentheses around '+'
> inside '<<' [-Wparentheses]
>
> I appreciate that these aren't bugs, and are just gcc being over
> enthusiastic in it's warnings (which I've asked it to be on the
> command line), however would it be worth sticking some parentheses
> into the code simply to avoid this kind of noise being visible for
> users?
>
> Regards,
> Andres
Try with -Wall -Wextra -Wpedantic. GCC will spam warnings for nearly 
anything, struct padding warnings are extremely annoying. Somewhat 
(maybe?) concerning warnings are signed / unsigned comparisons in math/*

Re: endian.h warnings

[Andre Renaud]

Hi Rich,
> I really don't understand why gcc is generating warnings for these; I
> thought it automatically suppressed warnings for system include
> directories, and even macro expansion text generated from macros
> defined in these headers. If that suppression is not working
> correctly, maybe we should work around these warnings. I agree it's
> problematic for users to get spammed with irrelevant warnings that
> decrease the signal to noise ratio of the compiler's warning
> generation.

You're right, this was caused because I wasn't correctly using the
musl-gcc wrapper script, so gcc wasn't aware that they were system
includes.

Regards,
Andre

Re: endian.h warnings

[Szabolcs Nagy]

* Nathan McSween <nwmcsween@gmail.com> [2013-07-09 19:29:12 -0700]:
> anything, struct padding warnings are extremely annoying. Somewhat
> (maybe?) concerning warnings are signed / unsigned comparisons in
> math/*

signed anything is concerning in math

(int arithmetics is used for bit representation manipulation
which should not be done signed, but this is a legacy from
fdlibm that will slowly get fixed)

i don't see signed comparision warnings with gcc,
i see a warning about the long double struct on x86
which is planned to be fixed but i wanted to wait until
i get around doing a larger long double code restructuring