Re: [MiNT] CVS Errors

[Eero Tamminen <oak@helsinkinet.fi>]

Hi,

On lauantai 08 tammikuu 2011, Thomas Juerges wrote:
> On 2011-01-08 13:00, Eero Tamminen wrote:
> > Solution to -Werror is not to use it. :-)
> 
> That is certainly the wrong solution.
> 
> Warnings are reported for a reason.  There is a probability > 0 that
> those warnings will be reported as errors in the future compiler
> releases. Warnings do not exist for the amusement of the crowd but to
> make the developer aware that the code causing the warning may or will
> cause unexpected results.  Use of the -Werror option should be mandatory
> and it is beyond me what the problem with fixing warnings is.

Valid warnings should be of course fixed.  What I'm saying is that -Werror
isn't necessarily the best way to *catch* them because it prevents builds
and shows you just the issues in the first C-file where the issues were
encountered.


Note that there can also be "unfixable" warnings:

* GCC generates sometimes bogus warnings (it's rare, but there are bugs
  about this in GCC bugzilla) and these differ between GCC versions

* Warning may come e.g. from a system header over which you don't have
  control

* Fixing a warning properly may be too much work (e.g. require redesign
  to propagate constness through everything)

* With portable software, different systems may trigger different warnings
  (from above three categories)

Sometimes it's best just to disable some warning, but usually I think it
better to leave & comment a warning from code with TODO than disable
some specific GCC compiler warning completely.


	- Eero