On 6/2/20 2:43 PM, Richard Shaw wrote:
On Tue, Jun 2, 2020 at 6:38 AM Panu Matilainen <pmati...@redhat.com <mailto:pmati...@redhat.com>> wrote:

    On 6/2/20 2:34 PM, Richard Shaw wrote:
     > On Tue, Jun 2, 2020 at 5:13 AM Panu Matilainen
    <pmati...@redhat.com <mailto:pmati...@redhat.com>
     > <mailto:pmati...@redhat.com <mailto:pmati...@redhat.com>>> wrote:
     >
     >
     >     Well, it's a guess.
     >
     >     Just tested the particular case of boost-devel install with plain
     >     rpm on
     >     all of sqlite, bdb and ndb, and of the three, bdb is the
    slowest one.
     >     The tested, expected behavior of sqlite is as fast or faster
    than bdb,
     >     but of course it's *possible* previously unknown worst-case
     >     behaviors exist.
     >
     >     Lets start with the basics:
     >     - is sqlite even involved - it will only be used on rawhide
    builds if
     >     mock bootstrap is used
     >     - does it make a difference if you override _db_backend to
    bdb/sqlite
     >     from mock config / cli define
     >     - a reproducer please (eg, what package is considerably
    slower to build
     >     than before, and by how much)
     >
     >
     >
     > boost-devel, nosync=True, bootstrap=False
     > real 0m6.083s
     > user 0m4.232s
     > sys 0m1.155s
     >
     > real 0m6.134s
     > user 0m4.356s
     > sys 0m1.115s
     > ---
     >
     > boost-devel, nosync=True, bootstrap=True
     > real 0m7.464s
     > user 0m4.849s
     > sys 0m1.364s
     >
     > I think this one was only slower because it converted the
    database? I
     > saw a bunch of messages about looking for sqlite but finding bdb
    database.
     >
     > real 0m7.234s
     > user 0m4.984s
     > sys 0m1.321s
     >
     > Nope, about the same on a second run.
     > ---
     >
     > boost-devel, nosync=False, bootstrap=False
     > real 0m7.017s
     > user 0m4.388s
     > sys 0m1.144s
     > ---
     >
     > boost-devel, nosync=False, bootstrap=True
     > real 1m13.294s
     > user 0m6.723s
     > sys 0m2.761s
     > ---
     >
     > So it looks like boostrap=True and nosync=False is the culprit,
    which I
     > inadvertantly got myself into. I did have either set and of course
     > nosync would be False by default and it looks like bootstrap=True by
     > default for rawhide.

    What's the actual command you're running here?


$ mock -r fedora-rawhide-x86_64 --init
$ time mock -r fedora-rawhide-x86_64 --install boost-devel

Ack, thanks.

So this nosync thing is some LD_PRELOAD hack. Is this actually used in our infra?

        - Panu -
_______________________________________________
devel mailing list -- devel@lists.fedoraproject.org
To unsubscribe send an email to devel-le...@lists.fedoraproject.org
Fedora Code of Conduct: 
https://docs.fedoraproject.org/en-US/project/code-of-conduct/
List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines
List Archives: 
https://lists.fedoraproject.org/archives/list/devel@lists.fedoraproject.org

Reply via email to