Re: Unsetting ENV in pkgsrc

[Tim Zingelman <tez%netbsd.org@localhost>]

On Sun, May 22, 2011 at 1:46 PM, Tim Zingelman <tez%netbsd.org@localhost> wrote:
> On Sun, May 22, 2011 at 1:35 PM, David Brownlee <abs%netbsd.org@localhost> 
> wrote:
>> On 22 May 2011 19:26, Tim Zingelman <tez%netbsd.org@localhost> wrote:
>>> On Sun, May 22, 2011 at 1:08 PM, David Brownlee <abs%netbsd.org@localhost> 
>>> wrote:
>>>> On 16 May 2011 15:21, Martin S. Weber <Ephaeton%gmx.net@localhost> wrote:
>>>>> On Sun, May 15, 2011 at 09:56:25AM +0200, Bernd Ernesti wrote:
>>>>>> So is ENV enough which has to be cleared and in which step should that
>>>>>> be done?
>>>>>
>>>>> I'm wondering why SHELL can't be "env -i sh" basically anyways...
>>>>
>>>> I've just tested adjusting the value of TOOLS_PLATFORM.env to include
>>>> " -i" and everything seems to build so far (about 800 packages into my
>>>> standard set on a fresh pkgsrc build), and it fixes my ENV issue.
>>>>
>>>> Would anyone object to appending -i to the TOOLS_PLATFORM.env
>>>> definitions in mk/tools/tools.*.mk ?
>>>
>>> Does bootstrap use these definitions?  If so I think that would break
>>> the current instructions for bootstrapping on several supported OS's.
>>> For instance in bootstrap/README.Solaris suggests:
>>>
>>>        env CC=/opt/studio12/SUNWspro/bin/cc \
>>>          SUNWSPROBASE=/opt/studio12/SUNWspro \
>>>          ./bootstrap --abi=64
>>
>> Is there someone using pkgsrc on Solaris who could test?
>> Just append change TOOLS_PLATFORM.env in mk/tools/tools.SunOS.mk to
>> "TOOLS_PLATFORM.env?=            /usr/bin/env -i" and run the bootstrap.
>>
>
> Trying to revive / update my test system now...

Bootstrap of HEAD fails on Solaris 10 with:

===> Building binary package for bootstrap-mk-files-20090807nb2
=> Creating binary package
/usr/pkgsrc/packages/All/bootstrap-mk-files-20090807nb2.tgz
pkg_create: Package building is not supported in bootstrap mode
*** Error code 1

unless you do 'env USE_DESTDIR=no ./bootstrap', and adding the -i to
TOOLS_PLATFORM.env causes this workaround to fail.

Let me see if I can hack around that trouble and try testing again.

 - Tim