[Mesa-dev] [PATCH 14/15] scons: Silence conversion from 'size_t' to 'type', possible loss of data on MSVC.

[Jose Fonseca]

On 20/03/15 14:36, Brian Paul wrote:
> On 03/20/2015 07:59 AM, Jose Fonseca wrote:
>> Most cases seem harmless, though that might not always be the case.
>> Maybe
>> one day we can get gcc to complain about these and fix them throughout
>> the code, but until then let's silence them.
>> ---
>>   scons/gallium.py | 1 +
>>   1 file changed, 1 insertion(+)
>>
>> diff --git a/scons/gallium.py b/scons/gallium.py
>> index 52f1b8d..bf6d605 100755
>> --- a/scons/gallium.py
>> +++ b/scons/gallium.py
>> @@ -507,6 +507,7 @@ def generate(env):
>>               '/wd4018', # signed/unsigned mismatch
>>               '/wd4056', # overflow in floating-point constant arithmetic
>>               '/wd4244', # conversion from 'type1' to 'type2',
>> possible loss of data
>> +            '/wd4267', # 'var' : conversion from 'size_t' to 'type',
>> possible loss of data
>>               '/wd4305', # truncation from 'type1' to 'type2'
>>               '/wd4351', # new behavior: elements of array 'array'
>> will be default initialized
>>               '/wd4756', # overflow in constant arithmetic
>>
>
> Would this silence cases of casting pointers to uint32, for example?  I
> don't think we'd want to silence that kind of warning.

I've just tested that, and it triggers a different warning -- C4047 ( 
https://msdn.microsoft.com/en-us/library/3t92260h.aspx ) --, which we 
still get.

> Anyway, the series looks good.
>
> Reviewed-by: Brian Paul <brianp at vmware.com>

Thanks.

With the series pretty much all MSVC C++ warnings (with /W3) are gone 
(thought I haven't tried all MSVC version and release type combinations.)


Left are linker warnings:

   ...: warning LNK4221: This object file does not define any previously 
undefined public symbols, so it will not be used by any link operation 
that consumes this library

which is totally bening, impossible to turn off, and hard to address, 
particularly for release builds where a lot of debug code gets turned off.

And

   warning LNK4197: export 'glClear' specified multiple times; using 
first specification
   ...

Which we should address -- either by stop using 
__declspec(dllimport/export) -- or stop using .DEFs.  My prefering would 
be to stop using __declspec, as .DEFs are good to spot bugs when symbols 
are missing.

Jose