[Swift-devel] duplicated job submission in swift-0.92?

[Michael Wilde]

Mihael,

My understanding is that Justin used a binary search approach: he kept extracting a point-in-time snapshot from SVN, and built and tested it, selecting dates in binary-search mode until he narrowed down the revision range that caused the failure to between r3835 and r3837.

Justin can confirm this when he sees this message, but don't count on confirmation very soon this weekend. Best to do you own tests to verify (or I can if that would help).

- Mike

----- Original Message -----
> That's possible, looking at the code.
> 
> Though what makes you think it's that?
> 
> Mihael
> 
> On Fri, 2011-04-01 at 17:33 -0500, Justin M Wozniak wrote:
> > I found that it appeared between Swift r3835 and r3837.
> >
> > On Fri, 1 Apr 2011, Michael Wilde wrote:
> >
> > > And Swift 0.91 works OK - it does *not* exhibit the twice-each
> > > bug.
> > >
> > > Justin: when you went backwards down the Swift 0.92 branch on
> > > Thursday
> > > morning, what did you find in terms of where it appeared the bug
> > > was
> > > introduced?
> > >
> > > - Mike
> > >
> > >
> > > ----- Original Message -----
> > >> I think we mist-spoke: The posted release 0.92 also exhibits the
> > >> twice-each bug as far as I acn tell.
> > >>
> > >> Mihael, Justin: can you test asap to confirm or refute that
> > >> observation?
> > >>
> > >> Thanks,
> > >>
> > >> - Mike
> > >>
> > >> which swift: ~/swift/rev/swift-0.92/bin/swift
> > >>
> > >> com$ swift -version
> > >> Swift svn swift-r4157 cog-r3056
> > >>
> > >> com$ cd ~/swift/lab
> > >> com$ cat zz3.swift
> > >> int arr[];
> > >>
> > >> arr[0]=1;
> > >> arr[1]=2;
> > >>
> > >> foreach a in arr {
> > >> trace("for", a);
> > >> }
> > >>
> > >> com$ swift zz3.swift
> > >> Swift svn swift-r4157 cog-r3056
> > >>
> > >> RunID: 20110401-1645-yyy87p39
> > >> Progress:
> > >> SwiftScript trace: for, 2
> > >> SwiftScript trace: for, 1
> > >> SwiftScript trace: for, 1
> > >> SwiftScript trace: for, 2
> > >> Final status:
> > >> com$
> > >>
> > >>
> > >> ----- Original Message -----
> > >>> I think both are good as they are.
> > >>>
> > >>> Would you like me to send it?
> > >>>
> > >>> Mihael
> > >>>
> > >>> On Thu, 2011-03-31 at 20:57 -0500, Michael Wilde wrote:
> > >>>> And I will send this to swift-user:
> > >>>>
> > >>>> "Dear Swift Users,
> > >>>>
> > >>>> On March 29 we discovered that the Release 0.92 branches of the
> > >>>> Swift and CoG trees were changed after the release and a
> > >>>> concurrency
> > >>>> bug was introduced. If you are running Swift from this *source
> > >>>> code*
> > >>>> base, please revert back to a known-working release such as the
> > >>>> 0.92
> > >>>> binary release if at all possible.
> > >>>>
> > >>>> We're working on restoring the 0.92 SVN branch to the correct
> > >>>> state
> > >>>> and will report back to this email list when that is done."
> > >>>>
> > >>>> Anything else to say? Feel free to send this out, adjusted as
> > >>>> needed, or just tell me what to change and I will.
> > >>>>
> > >>>> - Mike
> > >>>>
> > >>>>
> > >>>> ----- Original Message -----
> > >>>>> Please check this proposed warning on the Downloads page and
> > >>>>> let
> > >>>>> me
> > >>>>> know if its what we need there:
> > >>>>>
> > >>>>> http://www.ci.uchicago.edu/~wilde/swift/downloads/index.php
> > >>>>>
> > >>>>> I also fixed the 0.91 typo (but the downloads dont actually
> > >>>>> work
> > >>>>> from
> > >>>>> this test web. I think they will once this is committed and
> > >>>>> pushed
> > >>>>> live).
> > >>>>>
> > >>>>> - Mike
> > >>>>>
> > >>>>>
> > >>>>> ----- Original Message -----
> > >>>>>> ----- Original Message -----
> > >>>>>>> On Thu, 2011-03-31 at 20:14 -0500, Michael Wilde wrote:
> > >>>>>>>> ----- Original Message -----
> > >>>>>>>>> We decided the following:
> > >>>>>>>>> - I will revert the changes in the 0.92 branch
> > >>>>>>>>> - re-commit bug fixes that were committed after the
> > >>>>>>>>> merge
> > >>>>>>>>> - merge the 0.92 branch to trunk
> > >>>>>>>>> - fix the problems in trunk
> > >>>>>>>>
> > >>>>>>>> Sounds good. But when and how does the fix get to users?
> > >>>>>>>
> > >>>>>>> The package(s) are fine. Though we should probably also have
> > >>>>>>> a
> > >>>>>>> source
> > >>>>>>> package. The merge was done after the package(s) were
> > >>>>>>> uploaded
> > >>>>>>> to
> > >>>>>>> the
> > >>>>>>> swift site.
> > >>>>>>
> > >>>>>> Ah, great!
> > >>>>>>
> > >>>>>>> This only affects folks who have checked out from SVN the
> > >>>>>>> 0.92
> > >>>>>>> branch
> > >>>>>>> after the merge 9 days (or so) ago.
> > >>>>>>
> > >>>>>> Hmm - I question that. The release we use, based on 0.92 on
> > >>>>>> Beagle,
> > >>>>>> shows the twice-each error, and it was made on Feb 25, about
> > >>>>>> 35
> > >>>>>> days
> > >>>>>> ago. Does this merit clarification?
> > >>>>>>
> > >>>>>>> We should send an email to the user list once this is fixed.
> > >>>>>>> We
> > >>>>>>> may
> > >>>>>>> also
> > >>>>>>> want to send an email warning them not to check out from SVN
> > >>>>>>> but
> > >>>>>>> download the precompiled package instead.
> > >>>>>>
> > >>>>>> OK. I cant say that this will reach everyone. Perhaps some
> > >>>>>> status
> > >>>>>> notes on the Download page are in order. The 0.91 link there
> > >>>>>> is
> > >>>>>> wrong,
> > >>>>>> so we need to fix that page anyways.
> > >>>>>>
> > >>>>>>> I am a bit confused though. I would have expected the
> > >>>>>>> release
> > >>>>>>> to
> > >>>>>>> come
> > >>>>>>> with some announcement of some form.
> > >>>>>>
> > >>>>>> Agreed. We kept this low profile because we were trying to
> > >>>>>> coordinate
> > >>>>>> it with a Web change that we never accomplished. And we've
> > >>>>>> lost
> > >>>>>> the
> > >>>>>> habit of swift-user announcements but got to get back to
> > >>>>>> doing
> > >>>>>> that.
> > >>>>>> So, yes.
> > >>>>>>>
> > >>>>>>>>
> > >>>>>>>> Either create a 0.92.1 release (sounds hard based on
> > >>>>>>>> above)
> > >>>>>>>> or create a 0.93 release (in which case should we create
> > >>>>>>>> the
> > >>>>>>>> 0.93
> > >>>>>>>> branch from trunk as soon as this is fixed?)
> > >>>>>>>>
> > >>>>>>>> How long to re-test? (Thats a question for Sarah, Justin,
> > >>>>>>>> and
> > >>>>>>>> Ketan)
> > >>>>>>>> Could this include the Cray support mods?
> > >>>>>>>
> > >>>>>>> No! Fixing a problem is not a venue for introducing untested
> > >>>>>>> things
> > >>>>>>> into
> > >>>>>>> a release.
> > >>>>>>
> > >>>>>> I meant the Cray feature for 0.93 not 0.92.1
> > >>>>>> Yes, that should be tested.
> > >>>>>> But its being used pretty heavily.
> > >>>>>>
> > >>>>>> - Mike
> > >>>>>>
> > >>>>>>> But it could be discussed separately :)
> > >>>>>>>
> > >>>>>>> Mihael
> > >>>>>>
> > >>>>>> --
> > >>>>>> Michael Wilde
> > >>>>>> Computation Institute, University of Chicago
> > >>>>>> Mathematics and Computer Science Division
> > >>>>>> Argonne National Laboratory
> > >>>>>>
> > >>>>>> _______________________________________________
> > >>>>>> Swift-devel mailing list
> > >>>>>> Swift-devel at ci.uchicago.edu
> > >>>>>> http://mail.ci.uchicago.edu/mailman/listinfo/swift-devel
> > >>>>>
> > >>>>> --
> > >>>>> Michael Wilde
> > >>>>> Computation Institute, University of Chicago
> > >>>>> Mathematics and Computer Science Division
> > >>>>> Argonne National Laboratory
> > >>>>>
> > >>>>> _______________________________________________
> > >>>>> Swift-devel mailing list
> > >>>>> Swift-devel at ci.uchicago.edu
> > >>>>> http://mail.ci.uchicago.edu/mailman/listinfo/swift-devel
> > >>>>
> > >>
> > >> --
> > >> Michael Wilde
> > >> Computation Institute, University of Chicago
> > >> Mathematics and Computer Science Division
> > >> Argonne National Laboratory
> > >>
> > >> _______________________________________________
> > >> Swift-devel mailing list
> > >> Swift-devel at ci.uchicago.edu
> > >> http://mail.ci.uchicago.edu/mailman/listinfo/swift-devel
> > >
> > >
> >

-- 
Michael Wilde
Computation Institute, University of Chicago
Mathematics and Computer Science Division
Argonne National Laboratory