[eclipse-pmc] Seeking approval for compiler options modification

[Maxime Daniel <maxime_daniel@xxxxxxxxxx>]

https://bugs.eclipse.org/bugs/show_bug.cgi?id=170704

We would like to split the compiler option that commands the error level
for null reference issues into three different options.

The current OPTION_ReportNullReference would be replaced by the following
set:
OPTION_ReportNullReference
OPTION_ReportPotentialNullReference
OPTION_ReportRedundantNullCheck

This is a breaking change because, assuming that OPTION_ReportNullReference
is positioned to warning, the new options being left at their default value
(aka ignore), the second and third warnings of the code snippet below would
no more be reported:

void foo(Object p) {
Object l = null;
if (p == null) {} // taint p: it may be null
l.toString(); // warning 1: l can only be null
p.toString(); // warning 2: p may be null; no more present unless
OPTION_ReportPotentialNullReference is set to warning
if (l == null) {} // warning 3: l can only be null; no more present unless
OPTION_ReportRedundantNullCheck is set to warning
}

In fact, OPTION_ReportNullReference would then only control how cases that
would certainly raise an NPE at runtime are reported.

The value of doing so is to enable the user to focus on the most dangerous
and most certain errors amongst those that the null analysis reports. This
has enabled us to detect a very few such cases in Eclipse code (that have
been reported separately), amongst which none was a false alarm.

We would then be glad to had that change to 3.3.

Cheers,
Maxime DANIEL
Developer, Eclipse JDT Core
IBM Paris Laboratory
69, rue de la Vecquerie
44600 Saint-Nazaire
maxime_daniel@xxxxxxxxxx / +33 (0) 2 51 16 40 13 / tl 87 40 13 / FAX - 40
01 / Mob. +33 (0) 6 84 63 99 62
' It takes nine months to make a baby, and it takes focused concentration
to make great breakthroughs.' - Stever Robbins - in HBS Working Knowledge,
Jan 2005
'The easiest way to predict the future is to invent it.' - Alan Kay