[CRIU] [PATCH] make: protobuf -- Rework build procedure
Ruslan Kuprieiev
kupruser at gmail.com
Wed Jan 21 11:29:01 PST 2015
On 01/21/2015 09:13 PM, Cyrill Gorcunov wrote:
> On Wed, Jan 21, 2015 at 09:02:58PM +0200, Ruslan Kuprieiev wrote:
>> On 01/21/2015 08:53 PM, Cyrill Gorcunov wrote:
>>> On Wed, Jan 21, 2015 at 10:43:05AM -0800, Saied Kazemi wrote:
>>>> Thanks for your quick reply.
>>>> Yes, that solved the problem. The failure was confusing because I was
>>>> compiling for months without any issues even when using -j. Anyway, I
>>>> see that after commit "crit: add --format hex option", one has to install
>>>> libprotobuf-dev for make to work.
>>>> --Saied
>>>> P.S. Isn't having a symbolic link in the build system to a hard coded
>>>> pathname fragile?
>>> As to me, we have to ship own copy. Ruslan, does this file is something
>>> really internal to protobuf-devel?
>> What do you mean by "really internal"?
> I mean that it's subject for change and not backward compatible.
Oh, I see. Protobuf is made to be backward compatible, so, I guess,
everything
should be fine even if we copy descriptor.proto. Especially, because we
have a
unique number for our custom options, so it won't be taken by anyone in
the future.
>> It is needed to make protoc-C happy and compile opts.proto and everything
>> that uses it. We don't actually use it inside criu, but we use it in crit,
>> where everything
>> is being compiled properly with protoc, which is a bit smarter than
>> protoc-c, but still
>> requires descriptor.proto from protobuf-devel.
>>
>> But, i think, descriptor.proto is quite standalone and could be just copied
>> to
>> protobuf/google/protobuf. But do we really need to copy it?
>> Is having an additional protobuf-devel package really problematic?
> The thing is that it's not problematic at all, sure. Just might be more
> convenient I think.
I've tried to copy descriptor.proto to protobuf/google/protobuf and
everything
seems to work fine(build, test, install) without protobuf-devel installed.
So I'll send a patch soon.
Thanks =)
More information about the CRIU
mailing list