[Libreoffice-commits] core.git: Silence some conversion warnings

[Stephan Bergmann]

[putting the ML back into the loop]

On 08/24/2015 10:56 AM, Matteo Casalin wrote:
> On Mon, 24 Aug 2015 10:37:38 +0200
> Stephan Bergmann <sbergman at redhat.com> wrote:
>
>> On 08/21/2015 10:30 PM, Matteo Casalin wrote:
>>> commit eb4cbea657b9038c488f1b1bcf5107cc226a6681
>>> Author: Matteo Casalin <matteo.casalin at yahoo.com>
>>> Date:   Sun Aug 16 16:32:13 2015 +0200
>>>
>>>       Silence some conversion warnings
>>>
>>>       Change-Id: I676ed010576f3a24b193ffc6c28a319bcc5ac968
>>>
>> [...]
>>> diff --git a/store/source/storbase.hxx b/store/source/storbase.hxx
>>> index 395870f..78a51a1 100644
>>> --- a/store/source/storbase.hxx
>>> +++ b/store/source/storbase.hxx
>>> @@ -512,9 +512,9 @@ struct PageData
>>>        void guard (sal_uInt32 nAddr)
>>>        {
>>>            sal_uInt32 nCRC32 = 0;
>>> -        nCRC32 = rtl_crc32 (nCRC32, &m_aGuard.m_nMagic, sizeof(sal_uInt32));
>>> +        nCRC32 = rtl_crc32 (nCRC32, &m_aGuard.m_nMagic, static_cast<sal_uInt32>(sizeof(sal_uInt32)));
>>
>> What compiler/switch is it that warns there?  It does not look like that
>> particular compiler/switch is very useful.  sizeof(sal_uInt32) is a
>> known, small integer at compile time.  It should be apparent to both the
>> compiler and humans that it fits into the sal_uInt32 argument to
>> rtl_crc32 without needing the static_cast noise.
>
> That's GCC -Wconversion, which I'm using to chase harmful downcasts (e.g. sal_Int32 to sal_uInt16 for OUString length, there are still some of them here and there). The point is that some of these downcasts, although safe in themselves, introduce a lot of noise in the logs and make the real issues difficult to detect.
> If I remember correctly in this case the warning was probably only in the result of "theSize - sizeof(G)" some lines below, I just casted both call places for simmetry. If that's undesired I'll revert the cast for the "sizeof(sal_uInt32)" calls.

So you're actually finding problems with that?  Good to know---probably 
makes the resulting increase in noise more tolerable. ;)

Anyway, if any of those "obvious" casts are not actually needed by your 
toolchain to silence warnings, then yes, might be good if you just 
removed those again.