cmspack.c:(.text+0x276b): undefined reference to `_cmsHalf2Float' error in build-infra build ('old' build works OK)

[Martijn Verburg]

Hi Alan,

The sanity checking step is good to hear. We had a bit of an informal BOF
here at Devoxx about it after some labs using build-infra ran into these
types of build issues.

So thanks Lana for running these going forwards!

Cheers,
Martijn

On Thursday, 15 November 2012, Alan Bateman wrote:

> On 15/11/2012 05:01, Phil Race wrote:
>
>> On 11/14/12 3:23 PM, Kelly O'Hair wrote:
>>
>>> On Nov 14, 2012, at 2:03 PM, Phil Race wrote:
>>>
>>>  Ah .. I guess we (2d) broke that when we added the new file from LCMS
>>>> 2.4 to
>>>> FILES_c_unix.gmk but had no clue this other new top-level file existed.
>>>> Kelly said get Makefile changes reviewed but didn't mention file lists.
>>>>
>>> I stated that anything in the make/ directories needed to be run by us.
>>>
>>
>> Can I quote what you said .. its not quite so clear :-
>>
> I don't think it's worth spending time on this, instead we just need to
> get the message out again that we have to live with two build systems for a
> transitional period and that it's easy to break the one that you aren't
> using.
>
> Some people that push to jdk8/tl are using the old build system, some are
> using the new build system. The latter group have had their builds broken
> several times recently, almost always after changes from master were pulled
> down to jdk8/tl. There have been one or two local breakages too but they
> are easy to fix quickly, it's the changes from distant planets (jdk8/awt
> and jdk8/2d in particular) that hurt as the lag to get the fixes through is
> just too long.
>
> Lana Steuck does most of the integrations and sync up of the jdk8 forests,
> she's going to do sanity builds with the new build prior to integrating,
> and I think perhaps prior to pulling too, so that should at least reduce
> the instances of breakage until the transition is done and the old build is
> 5 foot under.
>
> -Alan
>