On 8/30/12 1:17 AM, Andrew Haley wrote:
On 08/29/2012 11:53 PM, Stuart Marks wrote:
On 8/29/12 8:48 AM, Andrew Hughes wrote:
But presumably [-Werror] would be removed when everything is warning free?
-Werror should not be the default for everyone building OpenJDK, who then
end up having to fix or workaround issues which are nothing to do with them.
It's easy enough for those who want to check we don't regress to turn on
-Werror throughout, once everything is warning free.
Once everything is warning free, -Werror *should* be enabled by default. That's
how we'll keep everything warning free.
If everything were warning free, I don't see how people could be blocked by
warnings that are unrelated to their current changes. This does happen today
with implicit compilation, but it shouldn't happen in the bright shining future
where there are no warnings. :-)
-Werror is probably OK for Java code, but not for HotSpot.
We GCC developers keep adding new warnings, so poor souls who keep
up-to-date with new Fedora (and other leading-edge) distributions tend
to be the first to discover problems. Thus is very awkward for people
who don't understand advanced C++ features.
Yes, absolutely, I had intended this to be about Java warnings only, not about
GCC warnings. More precisely, this discussion is about potentially setting
JAVAC_WARNINGS_FATAL=true in the makefiles, which obviously applies only to
Java code.
(And to be absolutely clear, Dan's warnings fix will not change any makefile
warnings settings. The setting of JAVAC_WARNINGS_FATAL=true may occur in some
hypothetical future changeset.)
Sorry for the confusion.
s'marks
