From: Qing Xu <[email protected]>
Fixed following compiler warning:
- drivers/regulator/max8925-regulator.c:269:51: warning:
'regulator_idx' may be used uninitialized in this function
[-Wmaybe-uninitialized]
Signed-off-by: Qing Xu <[email protected]>
---
drivers/regulator/max8925-regulator.c | 3 ++-
1 files changed, 2 insertions(+), 1 deletions(-)
diff --git a/drivers/regulator/max8925-regulator.c b/drivers/regulator/max8925-regulator.c
index 2b54979..187b29b 100644
--- a/drivers/regulator/max8925-regulator.c
+++ b/drivers/regulator/max8925-regulator.c
@@ -282,7 +282,8 @@ static int __devinit max8925_regulator_probe(struct platform_device *pdev)
struct max8925_regulator_info *ri;
struct resource *res;
struct regulator_dev *rdev;
- int i, regulator_idx;
+ int i;
+ int regulator_idx = 0;
res = platform_get_resource(pdev, IORESOURCE_REG, 0);
if (!res) {
--
1.7.0.4
On Thu, Nov 22, 2012 at 10:11:06AM +0800, Qing Xu wrote:
> - int i, regulator_idx;
> + int i;
> + int regulator_idx = 0;
This sort of fix is rarely good without some analysis as to why this is
a sensible initialisation to do, just unconditionally initialising may
be masking a real issue in the control flow which the compiler has
identified.
On 11/23/2012 09:41 AM, Mark Brown wrote:
> On Thu, Nov 22, 2012 at 10:11:06AM +0800, Qing Xu wrote:
>
>> - int i, regulator_idx;
>> + int i;
>> + int regulator_idx = 0;
> This sort of fix is rarely good without some analysis as to why this is
> a sensible initialisation to do, just unconditionally initialising may
> be masking a real issue in the control flow which the compiler has
> identified.
In my build environment, there is no such compiler warning. :(
Adding this patch is just want to avoid kbuild test robot's warning.
But, in fact, it is not necessary to initialize regulator_idx.
for (i = 0; i < ARRAY_SIZE(max8925_regulator_info); i++) {
ri = &max8925_regulator_info[i];
if (ri->vol_reg == res->start) {
****** if regulator_idx can not get a match "i" here, it will return
-EINVAL in below code
regulator_idx = i;
break;
}
}
if (i == ARRAY_SIZE(max8925_regulator_info)) {
dev_err(&pdev->dev, "Failed to find regulator %llu\n",
(unsigned long long)res->start);
return -EINVAL;
}
How to solve such compiler warning?
On Fri, Nov 23, 2012 at 10:27:12AM +0800, Qing Xu wrote:
> But, in fact, it is not necessary to initialize regulator_idx.
> for (i = 0; i < ARRAY_SIZE(max8925_regulator_info); i++) {
> ri = &max8925_regulator_info[i];
> if (ri->vol_reg == res->start) {
> ****** if regulator_idx can not get a match "i" here, it will return
> -EINVAL in below code
> regulator_idx = i;
> break;
> }
> }
> if (i == ARRAY_SIZE(max8925_regulator_info)) {
> dev_err(&pdev->dev, "Failed to find regulator %llu\n",
> (unsigned long long)res->start);
> return -EINVAL;
> }
> How to solve such compiler warning?
Typically by reporting a compiler bug, though sometimes in the process
of doing that one finds out that there's some non-obvious way in which
the code can break.
On 11/25/2012 01:55 AM, Mark Brown wrote:
> On Fri, Nov 23, 2012 at 10:27:12AM +0800, Qing Xu wrote:
>
>> But, in fact, it is not necessary to initialize regulator_idx.
>> for (i = 0; i < ARRAY_SIZE(max8925_regulator_info); i++) {
>> ri = &max8925_regulator_info[i];
>> if (ri->vol_reg == res->start) {
>> ****** if regulator_idx can not get a match "i" here, it will return
>> -EINVAL in below code
>> regulator_idx = i;
>> break;
>> }
>> }
>> if (i == ARRAY_SIZE(max8925_regulator_info)) {
>> dev_err(&pdev->dev, "Failed to find regulator %llu\n",
>> (unsigned long long)res->start);
>> return -EINVAL;
>> }
>> How to solve such compiler warning?
> Typically by reporting a compiler bug, though sometimes in the process
> of doing that one finds out that there's some non-obvious way in which
> the code can break.
It seems not like a compiler bug, its logic is:
for(...; i<xxx; ...) {
if (...) {
regulator_idx = i
break;
}
}
if (i == xxx)
return ERROR;
If regulator_idx can not get a matched "i" value, code will return ERROR.
But it seems that compiler can not do so complex judge.
And, I think the code is safe even if regulator_idx is not initialized, also
because of the "return ERROR" judge.