[R-pkg-team] Bug#908662: r-cran-etm: autopkgtest regression: saved results mismatch

[Andreas Tille]

Hi again,

I realised that you published a new version but the issue with the test
suite remains (partially):


Test test.etmCIF passed
--- tests.etm.Rout.save_        2019-07-17 05:55:30.013900516 +0000
+++ tests.etm.Rout_     2019-07-17 05:55:30.017900518 +0000
@@ -246,7 +246,7 @@
 116  0.9990 116.0 4.764e-07 0.99767 1.0000      1       0
 124  0.9990 124.0 4.764e-07 0.99767 1.0000      2       0
 164  0.9990 164.0 4.764e-07 0.99767 1.0000      0       0
-183  1.0000 183.0 1.337e-19 1.00000 1.0000      1       1
+183  1.0000 183.0 1.032e-19 1.00000 1.0000      1       1
 > 
 > ## gonna play a bit with the state names
 > dd <- sir.cont


I tend to ignore this for now but it would be really helpful
if you could solve this in a future release.

Kind regards

       Andreas.

On Tue, Oct 02, 2018 at 07:09:38AM +0200, Andreas Tille wrote:
> Hi,
> 
> it is definitely architecture dependand.  I think the solution would be
> to do a comparison tollerating this micro-diffs.
> 
> Kind regards
> 
>        Andreas.
> 
> On Mon, Oct 01, 2018 at 11:35:02PM +0200, Arthur Allignol wrote:
> > Hi, 
> > 
> > sorry for not getting back to you earlier. Totally forgot about it...
> > 
> > If I read the diff well, there is differences in the 19th decimal (e.g., 1.337e-19 vs 1.032e-19). Could these kind of rounding error be platform specific? The saved results were obtained on a Mac. I could find some time to fire up a VM and see but I would tend to ignore the problem. 
> > 
> > Should I take further actions?
> > 
> > Thanks for your work! and best,
> > Arthur
> > 
> > 
> > > On 12. Sep 2018, at 14:15, Andreas Tille <tille at debian.org> wrote:
> > > 
> > > Hi Arthur,
> > > 
> > > there was a bug report filed against the Debian packaged version of etm.
> > > It seems the saved results are not matching the result of the test.
> > > 
> > > Could you please comment on this - may be we are doing something wrong?
> > > 
> > > Kind regards
> > > 
> > >       Andreas.
> > > 
> > > 
> > > On Wed, Sep 12, 2018 at 12:26:13PM +0200, Graham Inggs wrote:
> > >> Source: r-cran-etm
> > >> Version: 1.0.4-2
> > >> User: debian-ci at lists.debian.org
> > >> Usertags: regression
> > >> 
> > >> Hi R Package Team
> > >> 
> > >> Now that #903672 has been resolved, we can see that r-cran-etm fails its
> > >> autopkgtest [1] with the following errors:
> > >> 
> > >> --- tests.etm.Rout.save_	2018-09-10 10:04:21.100595425 +0000
> > >> +++ tests.etm.Rout_	2018-09-10 10:04:21.104595447 +0000
> > >> @@ -246,7 +246,7 @@
> > >> 116  0.9990 116.0 4.764e-07 0.99767 1.0000      1       0
> > >> 124  0.9990 124.0 4.764e-07 0.99767 1.0000      2       0
> > >> 164  0.9990 164.0 4.764e-07 0.99767 1.0000      0       0
> > >> -183  1.0000 183.0 1.337e-19 1.00000 1.0000      1       1
> > >> +183  1.0000 183.0 1.032e-19 1.00000 1.0000      1       1
> > >>> 
> > >>> ## gonna play a bit with the state names
> > >>> dd <- sir.cont
> > >> @@ -694,7 +694,7 @@
> > >>  0.90344   35 1.193e-04 0.88204 0.92484     54       4
> > >>  0.96816   54 4.374e-05 0.95520 0.98113     15       2
> > >>  0.98894   77 1.668e-05 0.98094 0.99695      5       0
> > >> - 1.00000  460 1.016e-20 1.00000 1.00000      1       1
> > >> + 1.00000  460 2.253e-19 1.00000 1.00000      1       1
> > >> 
> > >> Transition 1 1
> > >>       P time      var   lower   upper n.risk n.event
> > >> autopkgtest [10:04:21]: test run-unit-test: -----------------------]
> > >> autopkgtest [10:04:21]: test run-unit-test:  - - - - - - - - - - results - -
> > >> - - - - - - - -
> > >> run-unit-test        FAIL non-zero exit status 1
> > >> 
> > >> Regards
> > >> Graham
> > >> 
> > >> 
> > >> [1] https://ci.debian.net/packages/r/r-cran-etm/unstable/amd64/
> > > 
> > > -- 
> > > http://fam-tille.de
> > 
> > 
> 
> -- 
> http://fam-tille.de
> 

-- 
http://fam-tille.de