On Sun, 2009-04-19 at 23:06 +0200, Øyvind Harboe wrote:
> Committed.
> 
> Note that treating warnings as errors may well have to stay
> non-default. Be prepared for a bit of flak...

Did you mean "on as default"?  I hope that to be the case, as I think
most developers do not report warnings - only errors.  I would even want
releases to have it enabled, as any warning that escapes QA reflects
something that was missed (or a potentially "unsupported" compiler).
I think those issues should be taken as seriously as "real" errors.

Since I prefer to see reports from builds using -Werror, I believe the
--disable-werror workaround should be used only when a patch will not be
immediately forthcoming to address the issue.  Further, that option can
be made mandatory for certain corner cases as a last resort. heck, the
configuration script could do it automatically (e.g. for releases).

To play my own devil's advocate, -Werror does not fit the premise that
the build system should "fail gracefully".  While I believe that
principle applies more to run-time behaviors, developers spend more time
"running" the build system.  Without abandoning it completely, I posit
that a more practical policy would be "fail fast and hard, but then
provide a means to gracefully work around the problems (if possible)."

In any case, I am sensitive to the issues involved and will be
interested to see the community's reaction.

> Since you're willing to address issues that are brought up here, I'm
> applying the patch to see if we can navigate some dire straits
> w/a few patches and then be out into more open water...

It has not been hard to deal with the problems so far, so I doubt there
will be anything that causes big problems for recent versions of gcc.
In my mind, the bigger question is that of general compatibility with
the pool of non-gcc compilers that are actively being used; are such
even being used in this community?

If it turns out there is more work than I anticipated, I am definitely
in favor of removing -Werror from the default settings, but I hope the
community instead concludes the benefits outweigh the short-term price
and will work together to resolve any remaining and future warnings.

Cheers,

Zach
_______________________________________________
Openocd-development mailing list
Openocd-development@lists.berlios.de
https://lists.berlios.de/mailman/listinfo/openocd-development

Reply via email to