Re: [Xen-devel] [PATCH v2] libxl: add support for yajl 2.x

[Roger Pau Monnà <roger.pau@xxxxxxxxxxxxx>]

2012/1/23 Ian Campbell <Ian.Campbell@xxxxxxxxxx>:
> On Mon, 2012-01-23 at 11:30 +0000, Roger Pau Monnà wrote:
>> 2012/1/23 Ian Campbell <Ian.Campbell@xxxxxxxxxx>:
>> > On Tue, 2012-01-03 at 18:00 +0000, Ian Jackson wrote:
>> >> Ian Campbell writes ("Re: [Xen-devel] [PATCH v2] libxl: add support for 
>> >> yajl 2.x"):
>> >> > On Thu, 2011-12-22 at 16:41 +0000, Roger Pau Monnà wrote:
>> >> > > Some autogoo stuff will be good, at least for the tools build. Can you
>> >> > > set up the basic structure Ian?
>> >> >
>> >> > I'm not going to have time in the near future.
>> >> >
>> >> > Would it be sufficient to have the stuff in tools/check create a
>> >> > config.h as well as doing the checks? Might need a bit of Makefile magic
>> >> > to be sure that check is always run?
>> >>
>> >> For now, I think we should not add more stuff to the critical path for
>> >> Xen 4.2. ÂBut after then we should switch to autoconf I think. Â(I
>> >> don't approve of automake.)
>> >
>> > It seems that yajl 2.0 support is blocked on the autoconf stuff. Can we
>> > disentangle things such that we can support yajl 2.0 in Xen 4.2 and
>> > switch to autoconf afterwards or do we have/want to make the switch to
>> > autoconf for 4.2?
>>
>> The proposed autoconf patch is just a convenient replacement for
>> tools/check scripts, which allows us to generate a makefile and a
>> header we can include to know the system capabilities, nothing more
>> was added.
>>
>> > I appreciate the concerns about critical path for 4.2. I presume
>> > autoconf would require more than just checking in the patches,
>> > specifically I'm thinking of the automated test system update and docs.
>>
>> Don't know how difficult it is to update the test bed to use the new
>> configure script, ideally it should only require adding "./configure"
>> before performing a "make tools". Since I don't know how the test bed
>> works, it might be necessary to pass some options to configure
>> execution to obtain the same result.
>
> Ian J would have to answer that one. Hopefully just running ./configure
> will get us as close as possible to the existing setup.
>
> FWIW the tester code is available at a git url which is in every posted
> set of results. Probably looking for anywhere that writes to .config
> would be a good start for deciding how far from the defaults it deviates
> (not far, I expect).
>
>> > On the other hand Roger posted v3 of his autoconf support patch and
>> > although I haven't got round to reviewing it yet (sorry) v2's review was
>> > generally positive/minor looking.
>>
>> v3 is basically a v2 with all the mentioned fixes and the output from
>> autoconf & automake, so we can use the configure script straight away.
>
> automake? I thought we agreed not to use that?

Autotools in general is quite a mess, we don't use automake, but we
need it to generate config.sub and config.guess (yes, I know I can
also copy them). That's all we need automake for (don't know why
autoconf can't do this automatically...)

>> Anyway, this patch for adding support of yajl 2.0 needs some rework,
>> according to the comments made by Ian Jackson, which I will try to do
>> before the end of the week (sorry for the delay, but I'm quite busy
>> this days).
>
> Sure, no problem. I have put yajl2 support in the "nice to have column"
> and autoconf in a new "need to decide if this is 4.2 or 4.3 material"
> column.

Thanks, I'm sure yajl 2.0 support will get into 4.2 :)

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel