Terminal newline test fails on non-english locale

[AshesITR]

The terminal newline detection mechanism uses grepl with a hard-coded english warning message. On systems with a different R language, this warning will not be caught.

Similarly to expect_lint we should temporarily set LANGUAGE=en during get_source_expressions.

[MichaelChirico]

We'll see if it passes on Windows, which is a pain:

It is usually possible to change the language once R is running via (not Windows) Sys.setlocale("LC_MESSAGES", "new_locale"), or by setting an environment variable such as LANGUAGE

https://cran.r-project.org/doc/manuals/R-admin.html#Localization-of-messages

[AshesITR]

I can say for sure the Sys.setenv(LANGUAGE = "en") approach works on my non-english Windows 10. I think the mentioned problem is just LC_MESSAGES being unsupported on Windows.

[MichaelChirico]

Please also test for cacheing effects:

https://github.com/r-lib/testthat/pull/1258#issuecomment-737546865

[AshesITR]

I'm unsure how to test and / or remedy them. Do you have a suggestion?

[MichaelChirico]

I think we can make a gettext approach work:

LANGUAGE=zh_CN R
> gettextf("incomplete final line found on '%s'", 1, domain="R")
# [1] "读'%s'时最后一行未遂"

The only complication is that it's a templated message

[AshesITR]

So you mean to grepl for whatever gettextf() returns? If yes, we should do the same in all other places where we currently use the LANGUAGE workaround.

[AshesITR]

I've just tried in a fresh R session:

gettextf("incomplete final line found on '%s'", 1, domain="R")
#> [1] "unvollständige letzte Zeile in '1' gefunden"
Sys.setenv(LANGUAGE = "en")
gettextf("incomplete final line found on '%s'", 1, domain="R")
#> [1] "incomplete final line found on '1'"
Sys.unsetenv("LANGUAGE")
gettextf("incomplete final line found on '%s'", 1, domain="R")
#> [1] "unvollständige letzte Zeile in '1' gefunden"

So there seems to be no caching issue, at least on R 4.0.3 @ Windows 10

[MichaelChirico]

is that windows too?

On Fri, Dec 4, 2020, 1:32 PM AshesITR notifications@github.com wrote:

I've just tried in a fresh R session:

gettextf("incomplete final line found on '%s'", 1, domain="R")

> [1] "unvollständige letzte Zeile in '1' gefunden"

Sys.setenv(LANGUAGE = "en")

gettextf("incomplete final line found on '%s'", 1, domain="R")

> [1] "incomplete final line found on '1'"

So there seems to be no caching issue, at least on R 4.0.3

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/jimhester/lintr/issues/661#issuecomment-738946431, or unsubscribe https://github.com/notifications/unsubscribe-auth/AB2BA5PTKV3GUUW55HSTSZDSTETKRANCNFSM4UNXCKMQ .

[AshesITR]

Yep.

[MichaelChirico]

OK great. Maybe one last test on an old version too?

I'm pretty apprehensive because this sort of issue has caused major headaches in the recent past:

https://github.com/Rdatatable/data.table/pull/4524

[AshesITR]

I don't have any old versions at hand atm. However, the issue you reference seems to be related to installed or available system languages.

On my system, I can get the error messages in french without a problem using LANGUAGE=fr, but no luck getting chinese output. Neither of cn, zh_CN, zh give a chinese message.

[MichaelChirico]

OK that makes more sense. presumably you'd have to have chinese language installed for the messages to appear? though maybe not if it's just doing a lookup of the env and matching it to the po file.

anyway I'm more confident we can proceed with set/unsetting the language