Re: manywarnings and -f options

[Simon Josefsson]

Eric Blake <address@hidden> writes:

> On 12/03/2011 09:00 AM, Simon Josefsson wrote:
>> What does -funit-at-a-time really do?  My gcc 4.4 manual says:
>> 
>> `-funit-at-a-time'
>>      This option is left for compatibility reasons. `-funit-at-a-time'
>>      has no effect, while `-fno-unit-at-a-time' implies
>>      `-fno-toplevel-reorder' and `-fno-section-anchors'.
>> 
>>      Enabled by default.
>
> That's the case for 4.4 and later.  But in gcc 4.3, it was not
> unconditionally enabled, and as I said earlier, at least coreutils ran
> into a situation where gcc 4.3. failed to compile at -Werror because
> -Wdisabled-optimization warned that -fno-unit-at-a-time was required,
> which warning turned into an error.
>
> At this point, gcc 4.3 is slowly phasing out; most Linux distros and
> Cygwin have moved on to newer compilers, where the problem is less
> likely to happen.

Agreed -- so couldn't the answer to this situation be "don't use
--enable-gcc-warnings"?  I realize we may run into a similar situation
in the future though, and I don't have a good idea on how to resolve
that.  We could deal with it on a case-by-case basis.

>> The manual for -ipa-pure-const says:
>> 
>> `-fipa-pure-const'
>>      Discover which functions are pure or constant.  Enabled by default
>>      at `-O' and higher.
>> 
>> That seems harmless and as far I understand would not modify code
>> generation.
>
> It doesn't modify code generation, but it DOES modify warning
> generation, and in a way that negatively interacts with libtool:
> http://debbugs.gnu.org/cgi/bugreport.cgi?bug=10197

I see this problem from time to time as well, but I believe it is really
a libtool problem that it hides messages, and not a problem with the
triggers of that problem.  It is especially problematic when building
code that has a lot of '#if PIC' or '#if !PIC' in it.

/Simon