On ti, 17 maalis 2020, Vít Ondruch wrote:
Dne 16. 03. 20 v 18:15 Kevin Fenzi napsal(a):
> On Mon, Mar 16, 2020 at 10:38:07AM +0000, Paul Howarth wrote:
>> On Mon, 16 Mar 2020 11:22:40 +0100
>> Vít Ondruch <vondruch(a)redhat.com> wrote:
>>> I always thought that one should not call `rpm` during rpmbuild.
>>> Nevertheless I am not sure what was the reason? Probably locking of
>>> RPM db? Can somebody elaborate?
>> It couldn't be guaranteed to work in the case that the buildroot was
>> populated using a different version of rpm that used a different
>> version of libdb. That's not an issue that crops up much these days as
>> libdb hasn't been version-updated for years (due to licensing issues
> Except that now we are going to move to sqlite... :)
> (See rpm 4.16 change posted today)
Good point. I have proposed this guideline to FPC:
Also, I'll ask removal of `%requires_eq`. This seems to be used just by
samba, so it should not cause too much breakage.
Please do not remove it. We need it critically in Samba case.
Anyway, it is strange that the autogenerated requires in samba would
be enough. There are already quite a lot of them in samba-dc:
It is a long story but cutting it short, libldb is tightly coupled to
Samba upstream and should not be deviated between the build and the
# Force using libldb version to be the same as build version
# Otherwise LDB modules will not be loaded and samba-tool will fail
# See bug 1507420
Auto-generated requires are not enough for that.
For some of the history see the thread at
/ Alexander Bokovoy
Sr. Principal Software Engineer
Security / Identity Management Engineering
Red Hat Limited, Finland