[R-pkg-team] Bug#1029310: [Help] Re: Bug#1029310: src:r-cran-bayesplot: fails to migrate to testing for too long: autopkgtest regression

[Nilesh Patra]

On 23 January 2023 3:06:58 pm IST, Andreas Tille <tille at debian.org> wrote:
>Am Sat, Jan 21, 2023 at 10:17:06PM +0530 schrieb Nilesh Patra:
>> On 1/21/23 21:59, Andreas Tille wrote:
>> > Am Sat, Jan 21, 2023 at 09:46:07PM +0530 schrieb Nilesh Patra:
>> > > On 1/21/23 16:56, Andreas Tille wrote:
>> > > > as per DebCI there are 15 autopkgtest failures which seem to be
>> > > > connected to ggplot2 API somehow.[1]  Interestingly Salsa CI[2] does not
>> > > > show this autopkgtest error neither can I reproduce the problem on my local
>> > > > pbuilder.
>> > > > 
>> > > > Has anyone some idea what might be wrong?  If not I might ask bug reporter
>> > > > for more info.
>> > > 
>> > > I can't repro it either.
>> > 
>> > Relieving to know that I did not missed any basic thing here. ;-)
>> > 
>> > > But the log you point to is a fresh failure (today).
>> > > Did not dig deeper, but did you take a look at the `diff` between log that passes and the one
>> > > which does not?
>> > 
>> > Not yet.  Where can I find those old logs?
>> 
>> You can take a diff betweek ci log you pointed to and the log on salsa CI/local system.
>> The debci runs with package versions in testing, so there might be a bump in some dependency.
>
>I did a diff between the r-cran-* packages in the failing log and the
>successful log: 
> [...]
>
>to me those diffs are locking not really spectacular regarding to
>the actual issue.

Right.

>I admit I'm running out of ideas but for the moment my last resort
>is to skip the 8 affected test, let the packages migrate to testing
>and revert skipping the tests afterwards.

I did find an issue[3] which states ggplot2 breaks bayesplot 1.9.0 and spews the exact same errors we observe in debci.
There's also an upstream issue with bayesplot installation (similar stuff again) where it works for some people but not for others, see[4].
But then I'm not sure as to why it works for me, you and salsa CI. For now I'll have to dismiss it by calling it black magic.

Your approach sounds fair. Maybe you should do what you wrote (disable tests, let it migrate, enable tests again)

>[1] https://ci.debian.net/packages/r/r-cran-bayesplot/testing/amd64/
>[2] https://ci.debian.net/data/autopkgtest/testing/amd64/r/r-cran-bayesplot/28139180/log.gz
[3]: https://github.com/tidyverse/ggplot2/blob/main/revdep/problems.md#bayesplot
[4]: 
https://github.com/stan-dev/bayesplot/issues/297
--
Best,
Nilesh