[libcxx] test fixes <locale>

[Steve MacKenzie]

Not sure I follow.
The code is testing <locale> money_get.  Agree that the last argument is 
an input-output argument.

Pseudo-code
long double ex;
do_get(..., ex); // money_get writes to ex
assert(ex == <expected_value>); // tests for expected value

The test is passing, so I am assuming that money_get is writing the 
expected value to 'ex'.

However, scan-build is flagging the test expression:
assert(ex == expected_value)
as a "garbage value", which is true of an uninitialized variable.

So, initializing the variable removes the "garbage", and it is my 
understanding that the last argument is used as a buffer and that the 
actual value passed in has no impact on what gets written to it as an 
out-param.

So that is my understanding, if that is wrong, please provide concrete 
guidance that corrects my misunderstanding and I'd be happy to 
investigate further.

Thanks,
Steve



On 10/22/2014 06:22 AM, Jonathan Roelofs wrote:
> Steve,
>
> IIRC, the last parameter of these get functions is supposed to be a write-only
> return argument. Did you diagnose why the lack of initialization is causing
> problems (because that's what should be fixed, not this test)?
>
>
> Cheers,
>
> Jon
>
>
> On 10/21/14 9:05 PM, Steve MacKenzie wrote:
>> My scan-build run against libc++ tests and headers continues.. up to and
>> including localization tests.
>>
>> The diff is applicable to libcxx/test/localization tests only, all fixes address
>> dead initialization.
>>
>> Three suspected scan-build false positive bugs were opened as
>> well:21326,21333,21335.
>>
>> Progress of the run is published here:
>> http://stevemac123.wordpress.com/static-analysis-run/
>>
>> Thanks,
>> Steve MacKenzie
>>
>>
>> _______________________________________________
>> cfe-commits mailing list
>> cfe-commits at cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
>>