> -----Original Message----- > From: lng-odp [mailto:lng-odp-boun...@lists.linaro.org] On Behalf Of ext > Anders Roxell > Sent: Thursday, April 30, 2015 12:21 PM > To: Nicolas Morey-Chaisemartin > Cc: lng-odp@lists.linaro.org > Subject: Re: [lng-odp] [PATCH] api: time: force time defines as ULL to > avoid computation overflows on 32bits systems > > On 2015-04-30 09:42, Nicolas Morey-Chaisemartin wrote: > > > > > > On 04/29/2015 08:43 PM, Mike Holmes wrote: > > > But this is in the public API so it may be used in an application and > it > > > could now change the size of a structure in memory. > > > > Maybe I'm missing something but I don't see how. This is only a literal > value. The time type used by the API is already on 64b. > > Just to clarify what I meat. > The test that failed was an application seen from the ODP perspective. > Through this change in ODP we change how the application compiles and > works, so this change passes the interface between the ODP and the > application. And that's why I think this is an API change. > > NOTE: This change wont affect all platforms (those defining ULONG as 64 > bits wont be affected), but it will affect other applications > on other platforms. > > Cheers, > Anders >
We can think it also as a bug fix. Application expects this to work... uint64_t time = 10*ODP_TIME_SEC; ... but without the change, application needs to do this to be portable for 32 bit ... uint64_t time = 10*(uint64_t)ODP_TIME_SEC; ... which was not intention of the API. It's on the border, but I think that API bug corrections like this are OK for the master branch. ODP_VERSION_API_MINOR is there for this kind of fixes. We could actually make it still more explicit that time is a uint64_t type (this time through api-next). #define ODP_TIME_SEC ((uint64_t)1000000000) -Petri _______________________________________________ lng-odp mailing list lng-odp@lists.linaro.org https://lists.linaro.org/mailman/listinfo/lng-odp
