svn commit: r364891 - in head: . release release/scripts

Ruslan Garipov ruslanngaripov at gmail.com
Fri Sep 11 09:13:47 UTC 2020


On September 9, 2020 6:51:18 PM Glen Barber <gjb at freebsd.org> wrote:

> On Wed, Sep 09, 2020 at 11:45:49AM +0500, Ruslan Garipov wrote:
>> On August 28, 2020 2:19:33 AM Glen Barber <gjb at FreeBSD.org> wrote:
>> 
>> > Author: gjb
>> > Date: Thu Aug 27 21:19:16 2020
>> > New Revision: 364891
>> > URL: https://svnweb.freebsd.org/changeset/base/364891
>> >
>> > Log:
>> >   Merge the projects/release-git branch to head.
>> >   This allows building 13.x from Git instead of Subversion.
>> I'm sorry for the late response (and for the bad first one I made on this
>> commit recently via Gmail web client).
>> 
>> My build machine is completely offline.  I provide it with fresh snapshots
>> of the source and ports tree manually -- via tarballs which I'm pulling on
>> a FreeBSD machine having Internet connection.  Therefore, neither
>> "/scratch", nor the chrooted environment for me don't require Internet
>> access (I fetch distfiles for offline usage too).
>> 
>> Can we somehow remove that requirement for the devel/git port?  For
>> example, if one has set all three {SRC,PORTS,DOC}_UPDATE_SKIP variables,
>> then release/release.sh doesn't require devel/git to exist on the host.
>> 
>> I really don't need git on my build machine (and don't need it for
>> building).
>> 
>
> Can you verify if the attached patch works for your use case?  You will
> need to set NOGIT=1 in your release.conf.
Partially.  It doesn't require devel/git to create the initial build 
environment, but it does before beginning cross-build in the 
extra_chroot_setup().

You check there does the NOGIT have zero length.  If NOGIT isn't empty, 
then release.sh will ($(which git) returns nothing for me) try to install 
devel/git port or package again.  And NOGIT isn't empty because in this 
function my release.conf was already sourced.  So I had removed "not" 
(``!'') on line #278, ignoring NOGIT checking, and cross-build started.

It failed eventually on modifications made by r365398, but that's another 
story.

Thanks again, Glen!

>
> Glen
>




More information about the svn-src-all mailing list