quickstep-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Harshad Deshmukh <hars...@cs.wisc.edu>
Subject Re: Quickstep Build failure on OSX 0.12 due to protobuf 2.7 deprecation warnings
Date Wed, 30 Nov 2016 15:31:10 GMT
+1 for upgrade from me too. Having richer data structure support like 
maps should be quite useful.

Do you have some idea about the changes we need to do in the proto 
files? I think Zuyu heavily uses proto for his work, so it will be 
useful to have his opinion on the upgrade.


On 11/29/2016 05:03 PM, Marc Spehlmann wrote:
> +1 for the upgrade option
>
> On Tue, Nov 29, 2016 at 11:45 AM, Saket Saurabh <ssaurabh@cs.wisc.edu>
> wrote:
>
>> Hi all,
>> I would like to start a discussion around how to fix the failing build
>> issue on OSX 10.12 due to deprecation of certain system APIs in the new OSX
>> version.
>> The Quickstep code does not rely on these deprecated APIs, but it is the
>> protobuf library which uses these deprecated APIs. [See:
>> https://github.com/google/protobuf/issues/2182]
>> This has already been resolved in the latest version of protobuf [Version
>>> = 3.0] [See: https://github.com/google/protobuf/pull/2337]
>> So, I foresee two possible ways to now fix this in Quickstep:
>>
>> 1.       Use the compile-time time flag -Wno-deprecated-declarations to
>> ensure that the deprecated warnings do not cause build time failure.
>>
>> a.       This may not be the best way to solve this. Turning off warnings
>> is always a bad idea, anyways.
>>
>> 2.       Upgrade to Protobuf Version 3.
>>
>> a.       This may require a one-time migration of our current .proto
>> files, but on the other hand, it will enable Quickstep to use the improved
>> Protobuf 3 library with richer APIs and features [e.g. maps, etc.].
>>
>> For now, Option 2 seems like a favorable long-term approach to solve this
>> problem. Kindly raise if someone has a concern with this approach, or has a
>> better alternative in mind.
>>
>> Thanks!
>> Saket
>>

-- 
Thanks,
Harshad


Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message