On 9/11/2013 1:12 AM, David Holmes wrote:
On 11/09/2013 3:19 PM, Joe Darcy wrote:
Hi David,
On 9/10/2013 8:31 PM, David Holmes wrote:
Hi Joe,
I assume these lint settings will cause the build to fail if someone
introduces a new warning.
That is the point actually :-)
I'm concerned that build breakage could easily be introduced. If that
happens is there a simple command-line make flag we can specify to
turn this off?
New code should be warning free for the lint categories which have been
cleared, either naturally or through judicious application of
@SuppressWarnings as needed. (For some other categories of warnings,
like deprecation, javac fixes are in-progress to improve the handling of
@SuppressWarnings.)
If you mean for experimentation, how can the settings be overridden,
then something like
make JAVAC_WARNINGS="-Xlint:-all"
should work.
I was concerned about "accidentally" pushing something that didn't go
through a proper build cycle. But that should be a rarity and of
course if the Java code compiles cleanly once then it will compile
cleanly always (unlike some other code :)).
As long as someone does a build using the makefiles in the repo, the
lint flags will get applied. If someone pushes without doing that much,
they need to be willing to accept the consequences of their actions ;-)
There is a bit of a hazard if the calculations of javac's warnings
change and a full build is not done beforehand; unexpected breakage
could result. However, the javac team is pretty good about doing extra
building when such a change is underway.
Thanks,
-Joe
