[openib-general] [PATCH 4/13] osm: port to WinIB stack : osmtest/osmtest.c

[Eitan Zahavi]

Hi Michael,

In general I agree we could make the code a little more safe by checking 
castings.
But in many of the cases (not the ones with user input - 
strtoul/strtoull) it is not required as the values are limited by the IB 
arch.

Anyway, the patch I am sending is for WinIB migration. Just doing the 
explicit cast does not make things any worst.
We could take the task of cleaning these integer casts (like I did in 
osm_pkey.c/h) but this is another patch.

EZ

Michael S. Tsirkin wrote:

>Quoting r. Eitan Zahavi <eitan at mellanox.co.il>:
>  
>
>>Subject: [PATCH 4/13] osm: port to WinIB stack : osmtest/osmtest.c
>>
>>Hi Hal
>>
>>Explicit cast required for the win compiler to handle this...
>>
>>Thanks
>>
>>Eitan
>>
>>Signed-off-by:  Eitan Zahavi <eitan at mellanox.co.il>
>>
>>Index: osmtest/osmtest.c
>>===================================================================
>>--- osmtest/osmtest.c	(revision 9502)
>>+++ osmtest/osmtest.c	(working copy)
>>@@ -3281,7 +3281,7 @@ osmtest_validate_path_data( IN osmtest_t
>>   else
>>   {
>>     /* Also, this doesn't detect fewer than the correct number of paths being returned */
>>-    if ( p_path->count >= ( 1 << lmc ) * ( 1 << lmc ) )
>>+    if ( p_path->count >= (uint32_t)( 1 << (2*lmc)) )
>>     {
>>       osm_log( &p_osmt->log, OSM_LOG_ERROR,
>>                "osmtest_validate_path_data: ERR 0052: "
>>
>>    
>>
>
>Integer casts are ugly, and can mask real errors.
>All you want is for the math result to be unsigned, so
> 1u << (2*lmc)
>would be cleaner with the same effect, I think.
>
>  
>