[R-pkg-devel] ICU init failed: U_FILE_ACCESS_ERROR

I think this error has to do with the binary version of stringi for
Windows oldrel being out-of-date in CRAN (version 1.1.7 while source is
at 1.2.4). The current version of stringi under Windows gives a
compilation note, so the binary is not available in CRAN. This causes
difficulties even with indirect dependencies (e.g. on lubridate). Even
lubridate itself is failing CRAN tests on Windows oldrel. Also one of my
packages is failing CRAN tests because it depends on lubridate. These
packages have been accepted by CRAN. The compilation note triggered when
building the stringi binary under Windows oldrel seems not to affect
stringi in use (circumstantial evidence rather than systematic checking)
and the current version of stringi can be installed from sources under
Windows oldrel and at least locally this seems to solve the type of
problems you are seeing.

Best regards,

Pedro.

On 2018-11-08 04:06, Derek Ogle wrote:

Ben ... I tried two more times today and still the same issue. Perhaps it is transient on a different time scale than what I have tested. I was hoping to not raise the R dependency level, but had thought about that as well. Thank you for your thoughts.

David ... thank you for the experience and advice. I will try noting that on my submission to CRAN and see what happens. I will report back so others can see my experience with this issue.

-----Original Message-----
From: R-package-devel <r-package-devel-bounces at r-project.org> On Behalf Of David Hugh-Jones
Sent: Wednesday, November 7, 2018 4:24 PM
To: Ben Bolker <bbolker at gmail.com>
Cc: r-package-devel at r-project.org
Subject: Re: [R-pkg-devel] ICU init failed: U_FILE_ACCESS_ERROR

My package had this issue and still got accepted. I think it is a known transient glitch. Just mention it when you submit.

On Wed, 7 Nov 2018 at 22:05, Ben Bolker <bbolker at gmail.com> wrote:

    Does it happen consistently?  If it's only happened once, could be a
transient glitch in package dependencies.  I'd try re-testing, as a
start.  Alternatively, if you're willing to add an R >= 3.5.1
dependency to your package, presumably CRAN won't mind if it fails
tests on old-release ...



On 2018-11-07 4:55 p.m., Derek Ogle wrote:

I am considering a CRAN release of a new package and was using

R-winbuilder as a check. The check was successful with all but the
"old-release" version (see
https://win-builder.r-project.org/UhyHnNyn4Ukz/00check.log). The two
errors both appear to be related to tests using
testhtat::expect_output() and begin with the "ICU init failed: U_FILE_ACCESS_ERROR" error message.
The full output for the first error is pasted below. My internet
searches suggest (to me, but I am not confident) that this error is
related to the stringr or stringi packages, which both appear to be
used by testthat::expect_output(). Given that I don't receive this
error in any other checking, I am not sure how to troubleshoot it
(besides simply removing this test). Any thoughts would be greatly appreciated. Thank you.

** running tests for arch 'i386' ... [5s] ERROR
    Running 'test-all.R' [4s]
Running the tests in 'tests/test-all.R' failed.
Complete output:

    > library(testthat)
    > test_check("RFishBC")

    Loading required package: RFishBC
    ## RFishBC v0.0.13.9000. See vignettes at derekogle.com/RFishBC/.

    -- 1. Error: backCalc() output types (@test_OUTPUTS.R#272)

--------------------

    ICU init failed: U_FILE_ACCESS_ERROR
    1: expect_output(backCalc(SMBassWB, lencap, BCM = "DALE", inFormat
=

"wide", digits = 0),

           "Dahl-Lea") at testthat/test_OUTPUTS.R:272
    2: quasi_capture(enquo(object), capture_output, label = label)
    3: capture(act$val <- eval_bare(get_expr(quo), get_env(quo)))
    4: capture_output_lines(code, print, width = width)
    5: eval_with_output(code, print = print, width = width)
    6: withr::with_output_sink(temp, withVisible(code))
    7: force(code)
    8: withVisible(code)
    9: eval_bare(get_expr(quo), get_env(quo))
    10: backCalc(SMBassWB, lencap, BCM = "DALE", inFormat = "wide",
digits

= 0)

    11: stringr::str_replace_all
    12: getExportedValue(pkg, name)
    13: asNamespace(ns)
    14: getNamespace(ns)
    15: tryCatch(loadNamespace(name), error = function(e) stop(e))
    16: tryCatchList(expr, classes, parentenv, handlers)
    17: tryCatchOne(expr, names, parentenv, handlers[[1L]])
    18: value[[3L]](cond)

--
Sent from Gmail Mobile

	[[alternative HTML version deleted]]