On Thu, Sep 10, 2015 at 11:10:22AM +0800, Fengguang Wu wrote:
Hi Takashi,
Fengguang, we've been discussing about the compile warnings that weren't caught by 0days. It seems that it's triggered by !CONFIG_OF but with CONFIG_COMPILE_TEST=y. Then I got warnings like:
sound/soc/codecs/cs35l32.c: In function ‘cs35l32_i2c_probe’: sound/soc/codecs/cs35l32.c:278:2: warning: ‘val’ may be used uninitialized in this function [-Wmaybe-uninitialized] switch (val) { ^ sound/soc/codecs/cs35l32.c:272:15: note: ‘val’ was declared here unsigned int val; ^
The above was with gcc-5.1.1, but Lars told that he saw such a warning with gcc-4.9, too.
Could you add this kind of kconfig in your test?
It's covered through the lots of randconfig tests. However the problem is, they are pretty old warnings and 0day ignores old warnings because old warnings may well be intensionally ignored by people. On the contrast, build errors can be re-reported if remain unfixed for long time.
That said, if as a maintainer you demand "all warnings should be quieted", I'll happily help you reminding the people who break the rule.
If it's a generally agreed rule by the maintainers, I'll be happy to help guarantee it kernel wide.
Thanks, Fengguang