Bug#927335: r-base breaks 9 autopkgtests

classic Classic list List threaded Threaded
5 messages Options
Reply | Threaded
Open this post in threaded view
|

Bug#927335: r-base breaks 9 autopkgtests

Paul Gevers-4
Source: r-base
Severity: important
User: [hidden email]
Usertags: breaks
Control: affects -1 r-cran-permute
Control: affects -1 r-cran-phangorn
Control: affects -1 r-cran-popepi
Control: affects -1 r-cran-recipes
Control: affects -1 r-cran-sp
Control: affects -1 r-cran-spam
Control: affects -1 r-cran-units
Control: affects -1 r-cran-vegan
Control: affects -1 r-cran-zelig

[X-Debbugs-CC: [hidden email],
[hidden email]]

Dear maintainers,

With a recent upload of r-base the autopkgtest of 9 r-cran-* fail in
testing (and unstable) when their autopkgtests are run with the binary
packages of r-base from unstable. It passes in testing when run with
only packages from testing.

Apart from the freeze, this regression is blocking the migration of
r-base to testing [1]. Can you please investigate the situation? If the
issue (also) needs to be solved in the packages with the regressing
autopkgtest than please clone this bug and reassign to the right
package. Because this uploads breaks so many autopkgtests, I filed this
bug only against r-base, normally I file these kind of bugs against both
packages that are involved, but I X-Debbugs-CC-ed the maintainers of the
other packages.

More information about this bug and the reason for filing it can be found on
https://wiki.debian.org/ContinuousIntegration/RegressionEmailInformation

Paul

[1] https://qa.debian.org/excuses.php?package=r-base


signature.asc (499 bytes) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Bug#927335: r-base breaks 9 autopkgtests

Andreas Tille-2
On Thu, Apr 18, 2019 at 09:28:55AM +0200, Paul Gevers wrote:

> Source: r-base
> Severity: important
> User: [hidden email]
> Usertags: breaks
> Control: affects -1 r-cran-permute
> Control: affects -1 r-cran-phangorn
> Control: affects -1 r-cran-popepi
> Control: affects -1 r-cran-recipes
> Control: affects -1 r-cran-sp
> Control: affects -1 r-cran-spam
> Control: affects -1 r-cran-units
> Control: affects -1 r-cran-vegan
> Control: affects -1 r-cran-zelig
>
> [X-Debbugs-CC: [hidden email],
> [hidden email]]
>
> Dear maintainers,
>
> With a recent upload of r-base the autopkgtest of 9 r-cran-* fail in
> testing (and unstable) when their autopkgtests are run with the binary
> packages of r-base from unstable. It passes in testing when run with
> only packages from testing.
>
> Apart from the freeze, this regression is blocking the migration of
> r-base to testing [1].

Well, "apart from freeze" is the keyword here.  The problem is that the
r-base maintainer has a long record of ignoring freeze policy but the
r-pkg team is not updating packages to new upstream versions in freeze
time.  So this will be fixed *after* the Stretch release when we start
to refresh all R packages with new upstream versions which we stopped
in Freeze time.

> Can you please investigate the situation?

I'm not motivated to react if others fail to understand freeze policy,
sorry.

> If the
> issue (also) needs to be solved in the packages with the regressing
> autopkgtest than please clone this bug and reassign to the right
> package. Because this uploads breaks so many autopkgtests, I filed this
> bug only against r-base, normally I file these kind of bugs against both
> packages that are involved, but I X-Debbugs-CC-ed the maintainers of the
> other packages.

Thanks for your work for autopkgtests in general

       Andreas.

--
http://fam-tille.de

Reply | Threaded
Open this post in threaded view
|

Bug#927335: r-base breaks 9 autopkgtests

Dirk Eddelbuettel
In reply to this post by Paul Gevers-4

On 18 April 2019 at 09:28, Paul Gevers wrote:
| Source: r-base
| Severity: important
| User: [hidden email]
| Usertags: breaks
| Control: affects -1 r-cran-permute
| Control: affects -1 r-cran-phangorn
| Control: affects -1 r-cran-popepi
| Control: affects -1 r-cran-recipes
| Control: affects -1 r-cran-sp
| Control: affects -1 r-cran-spam
| Control: affects -1 r-cran-units
| Control: affects -1 r-cran-vegan
| Control: affects -1 r-cran-zelig

I have argues before that these need to be reassigned as these packages
generally build at CRAN -- and test in a wider (OS-wise speaking) setting
than we have. See eg
  https://cloud.r-project.org/web/checks/check_results_permute.html
which gets OK on each of the 12 platforms -- across multiple releases and OSs.  

I have in the past tried to argue and explain how our system of autopkgtest
was in my eyes not the right approach _as it ignores the system R itself has
and uses_ and these repeated bug reports are just repurcussions of it. I am
tired of rehashing it as it goes nowhere.

They must be fixed in the respective packages.  R releases once a year with 3
to 4 minor releases, and I tend to upload thoses.  R 3.6.0 comes out April 26.
I plan to upload it. The Debian release can use the testing version as in the
past.

Dirk
 
| [X-Debbugs-CC: [hidden email],
| [hidden email]]
|
| Dear maintainers,
|
| With a recent upload of r-base the autopkgtest of 9 r-cran-* fail in
| testing (and unstable) when their autopkgtests are run with the binary
| packages of r-base from unstable. It passes in testing when run with
| only packages from testing.
|
| Apart from the freeze, this regression is blocking the migration of
| r-base to testing [1]. Can you please investigate the situation? If the
| issue (also) needs to be solved in the packages with the regressing
| autopkgtest than please clone this bug and reassign to the right
| package. Because this uploads breaks so many autopkgtests, I filed this
| bug only against r-base, normally I file these kind of bugs against both
| packages that are involved, but I X-Debbugs-CC-ed the maintainers of the
| other packages.
|
| More information about this bug and the reason for filing it can be found on
| https://wiki.debian.org/ContinuousIntegration/RegressionEmailInformation
|
| Paul
|
| [1] https://qa.debian.org/excuses.php?package=r-base
|
| x[DELETED ATTACHMENT signature.asc, application/pgp-signature]

--
http://dirk.eddelbuettel.com | @eddelbuettel | [hidden email]

Reply | Threaded
Open this post in threaded view
|

Bug#927335: r-base breaks 9 autopkgtests

Paul Gevers-4
Hi Dirk,

On Thu, 18 Apr 2019 05:38:34 -0500 Dirk Eddelbuettel <[hidden email]> wrote:
> I have argues before that these need to be reassigned as these packages
> generally build at CRAN -- and test in a wider (OS-wise speaking) setting
> than we have. See eg
>   https://cloud.r-project.org/web/checks/check_results_permute.html
> which gets OK on each of the 12 platforms -- across multiple releases and OSs.  

That check is for a version of r-cran-permute not yet in Debian, for
reasons explained by Andreas. I can't keep track of all ecosystems, so I
am just responding like I do to *all* autopkgtest regressions. We're in
a freeze to release a new Debian version. Fixes need to go through
unstable as much as possible, so new uploads there that don't target
buster can be harmful.

> I have in the past tried to argue and explain how our system of autopkgtest
> was in my eyes not the right approach _as it ignores the system R itself has
> and uses_ and these repeated bug reports are just repurcussions of it. I am
> tired of rehashing it as it goes nowhere.

Maybe r is different in Debian than other packages and/or ecosystems,
maybe not. But I'll stop helping you with autopkgtest regressions in the
r part of the archive from now on.

> They must be fixed in the respective packages.  R releases once a year with 3
> to 4 minor releases, and I tend to upload thoses.  R 3.6.0 comes out April 26.
> I plan to upload it. The Debian release can use the testing version as in the
> past.

Please avoid this next release cycle. Uploading new upstream releases
during the freeze isn't making releasing Debian easier and quite often
is harmful.

Paul


signature.asc (499 bytes) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Bug#927335: r-base breaks 9 autopkgtests

Dirk Eddelbuettel

On 18 April 2019 at 13:27, Paul Gevers wrote:
| That check is for a version of r-cran-permute not yet in Debian, for

A bug in r-cran-permute.

In Google-speak, 'CRAN lives at HEAD'. The top of its release repo (ie the
CRAN mirrors) is what we would call a release and guaranteed to work across
package. So one has to keep the repo consistent and current.

Some maintainers here tend not to do that; it is then there a problem in
their packages. I said my piece there and won't expand.

| We're in a freeze to release a new Debian version.

Please do what we have done with r-base for 20+ years: tag it as 'do not
migrate'.

| But I'll stop helping you with autopkgtest regressions in the
| r part of the archive from now on.

These are not mine. I consider these to be ill-designed out of possible
unfamiliarity with the R ecosystem. I don't know. I do not enable them in my
r-cran-* packages.

If this stuff breaks for someone else, "they get to keep the pieces".

Dirk

--
http://dirk.eddelbuettel.com | @eddelbuettel | [hidden email]