[Debian-med-packaging] Unit tests of R BatchJobs
Bernd Bischl
bernd_bischl at gmx.net
Thu Jun 26 20:31:00 UTC 2014
Hi Andreas,
1) I have added Michel to CC.
2) Thx, this is good news.
3) I need to look into this a bit closer, which I probably cannot do
until end of next week.
(teaching load + conference paper)
But I can say that we will support you guys in removing these problems
soon.
4) I will add up a issue in our github tracker so we can manage this there.
A few comments from just glancing at stuff:
5) We do not write the the temp dir for the unit tests (actually I think
we did in the past),
because we like to be able to run them on our true custers in full batch
mode as well, to detect more errors.
And on these systems, TEMP is usually not on a shared filesystem, so
this does not work.
But this can sure be made configureable in some way. I mainly need to
figure out whether we write to that dir in
some other place as well. The error you see is caused by just one line
of code that is reused everywhere through a unit test
helper function.
6) For the SQLite: I have no idea. Are you sure that this is due to any
code we control?
Best,
Bermd
On 26.06.2014 22:20, Andreas Tille wrote:
> Hi,
>
> the Debian Med team intends to package BatchJobs for Debian. Since we
> try to provide means for testing packages in so called autopkgtest I
> was running
>
> LC_ALL=C R --no-save < run-all.R
>
> and noticed that when doing this as normal user it results in
>
> 1. Error: batchExpandGrid ------------------------------------------------------
> Could not create dir: unittests-files/unittestee2f8ba97c/registry
> 1: makeTestRegistry() at test_batchExpandGrid.R:4
> 2: makeRegistry(id = "unittests", seed = 1, packages = packages, file.dir = rd, work.dir = "unittests-files",
> ...) at /usr/lib/R/site-library/BatchJobs/tests/helpers.R:39
> 3: makeRegistryInternal(id, file.dir, sharding, work.dir, multiple.result.files, seed,
> packages, src.dirs, src.files)
> 4: checkDir(file.dir, create = TRUE, check.empty = TRUE, check.posix = TRUE, msg = TRUE)
> 5: stop("Could not create dir: ", path)
>
> 2. Error: batchMap -------------------------------------------------------------
> Could not create dir: unittests-files/unittestee26666541a/registry
> 1: makeTestRegistry() at test_batchMap.R:4
> 2: makeRegistry(id = "unittests", seed = 1, packages = packages, file.dir = rd, work.dir = "unittests-files",
> ...) at /usr/lib/R/site-library/BatchJobs/tests/helpers.R:39
> 3: makeRegistryInternal(id, file.dir, sharding, work.dir, multiple.result.files, seed,
> packages, src.dirs, src.files)
> 4: checkDir(file.dir, create = TRUE, check.empty = TRUE, check.posix = TRUE, msg = TRUE)
>
>
> since normal users can not write to /usr/lib/R/site-library/.
>
> I tried as root which seems to work without error but I get
>
>> test_package("BatchJobs")
> batchExpandGrid : Warning in sqliteInitDriver(max.con, fetch.default.rec, force.reload, shared.cache) :
> RS-DBI driver warning: (SQLite mismatch between compiled version 3.8.5 and runtime version 3.8.4.3)
> Warning in sqliteInitDriver(max.con, fetch.default.rec, force.reload, shared.cache) :
> RS-DBI driver warning: (SQLite mismatch between compiled version 3.8.5 and runtime version 3.8.4.3)
> Warning in sqliteInitDriver(max.con, fetch.default.rec, force.reload, shared.cache) :
> RS-DBI driver warning: (SQLite mismatch between compiled version 3.8.5 and runtime version 3.8.4.3)
> .Warning in sqliteInitDriver(max.con, fetch.default.rec, force.reload, shared.cache) :
> RS-DBI driver warning: (SQLite mismatch between compiled version 3.8.5 and runtime version 3.8.4.3)
>
>
> Do you see any chance to let the test also work as normal user by
> writing to a temporary directory and also stop the warning about the
> SQLite version?
>
> Kind regards and thanks for providing BatchJobs as free software
>
> Andreas.
>
--
Currently head of working group "Computational Statistics", LMU Munich:
http://www.statistik.lmu.de/institut/ag/compstat/
My old page at TU Dortmund:
http://www.statistik.tu-dortmund.de/bischl.html
More information about the Debian-med-packaging
mailing list