Regression from 0.12.7: Incorrect message returned by stop(), on Travis only

[krlmlr]

Rcpp::stop() seems to be unable to propagate the error message, instead it always uses "basic_string::resize" as message.

I was unable to replicate this locally (on Ubuntu 16.10) nor in a fresh "precise" container with R 3.2.5 from deb http://cran.rstudio.com/bin/linux/ubuntu precise/.

Failing Travis run of a MWE: https://travis-ci.org/krlmlr/Rcpp.stop.test/builds/177795286#L1314

Successful Travis run, differs only by installing Rcpp 0.12.7: https://travis-ci.org/krlmlr/Rcpp.stop.test/builds/177800745

Perhaps related: #593.

CC @jimhester.

[eddelbuettel]

That isn't really reproducible.

However, you can start from Docker with Ubuntu 12.04, 14.04, 16.04, ... and install R and Rcpp and test. I pretty much had to do the same to build myself .deb packages of an upstream library current in Debian but not in 14.04 (which is what I use for Travis) or 16.04 (most of my servers).

[krlmlr]

That's the level of reproducibility I can afford to provide at this time.

I already did what you suggested but couldn't replicate the failure.

[eddelbuettel]

Just revived a wily container with R, re-installed Rcpp 0.12.8 -- it works. This may be /very/ tricky to catch, but then again if it doesn't bite anybody else anywhere else then .... well, do we really care?

[eddelbuettel]

And I can replicate it on precise:

~$ docker run --rm -ti rocker/r-apt:precise 
root@748b02fe620c:/# R

R version 3.2.5 (2016-04-14) -- "Very, Very Secure Dishes"
Copyright (C) 2016 The R Foundation for Statistical Computing
Platform: x86_64-pc-linux-gnu (64-bit)

R is free software and comes with ABSOLUTELY NO WARRANTY.
You are welcome to redistribute it under certain conditions.
Type 'license()' or 'licence()' for distribution details.

  Natural language support but running in an English locale

R is a collaborative project with many contributors.
Type 'contributors()' for more information and
'citation()' on how to cite R or R packages in publications.

Type 'demo()' for some demos, 'help()' for on-line help, or
'help.start()' for an HTML browser interface to help.
Type 'q()' to quit R.

> install.packages("Rcpp"); library(Rcpp); cppFunction("double foo(double x) { if (x > 10) stop(\"Oh noes\"); return x+x; }"); foo(12)
Installing package into ‘/usr/local/lib/R/site-library’
(as ‘lib’ is unspecified)
--2016-11-21 22:53:34--  https://cran.rstudio.com/src/contrib/Rcpp_0.12.8.tar.gz
Resolving cran.rstudio.com (cran.rstudio.com)... 52.84.2.8
Connecting to cran.rstudio.com (cran.rstudio.com)|52.84.2.8|:443... connected.
HTTP request sent, awaiting response... 200 OK
Length: 2435852 (2.3M) [application/x-gzip]
Saving to: `/tmp/RtmpsyVq8e/downloaded_packages/Rcpp_0.12.8.tar.gz'

100%[==========================================================================================================================================================================>] 2,435,852   11.1M/s   in 0.2s    

2016-11-21 22:53:34 (11.1 MB/s) - `/tmp/RtmpsyVq8e/downloaded_packages/Rcpp_0.12.8.tar.gz' saved [2435852/2435852]

* installing *source* package ‘Rcpp’ ...
** package ‘Rcpp’ successfully unpacked and MD5 sums checked
** libs
g++ -I/usr/share/R/include -DNDEBUG -I../inst/include/     -fpic  -g -O2 -fstack-protector --param=ssp-buffer-size=4 -Wformat -Wformat-security -Werror=format-security -D_FORTIFY_SOURCE=2 -g  -c Date.cpp -o Date.o
g++ -I/usr/share/R/include -DNDEBUG -I../inst/include/     -fpic  -g -O2 -fstack-protector --param=ssp-buffer-size=4 -Wformat -Wformat-security -Werror=format-security -D_FORTIFY_SOURCE=2 -g  -c Module.cpp -o Module.o
g++ -I/usr/share/R/include -DNDEBUG -I../inst/include/     -fpic  -g -O2 -fstack-protector --param=ssp-buffer-size=4 -Wformat -Wformat-security -Werror=format-security -D_FORTIFY_SOURCE=2 -g  -c Rcpp_init.cpp -o Rcpp_init.o
g++ -I/usr/share/R/include -DNDEBUG -I../inst/include/     -fpic  -g -O2 -fstack-protector --param=ssp-buffer-size=4 -Wformat -Wformat-security -Werror=format-security -D_FORTIFY_SOURCE=2 -g  -c api.cpp -o api.o
g++ -I/usr/share/R/include -DNDEBUG -I../inst/include/     -fpic  -g -O2 -fstack-protector --param=ssp-buffer-size=4 -Wformat -Wformat-security -Werror=format-security -D_FORTIFY_SOURCE=2 -g  -c attributes.cpp -o attributes.o
g++ -I/usr/share/R/include -DNDEBUG -I../inst/include/     -fpic  -g -O2 -fstack-protector --param=ssp-buffer-size=4 -Wformat -Wformat-security -Werror=format-security -D_FORTIFY_SOURCE=2 -g  -c barrier.cpp -o barrier.o
g++ -shared -L/usr/lib/R/lib -Wl,-Bsymbolic-functions -Wl,-z,relro -o Rcpp.so Date.o Module.o Rcpp_init.o api.o attributes.o barrier.o -L/usr/lib/R/lib -lR
installing to /usr/local/lib/R/site-library/Rcpp/libs
** R
** inst
** preparing package for lazy loading
** help
*** installing help indices
** building package indices
** installing vignettes
** testing if installed package can be loaded
* DONE (Rcpp)

The downloaded source packages are in
        ‘/tmp/RtmpsyVq8e/downloaded_packages’
Error in foo(12) : basic_string::resize
> 

That is probably an interaction with the compiler version.

[eddelbuettel]

Ditto on precise aka 14.04

[eddelbuettel]

So two out of four fail, but those are old. Not sure how much I care. Per this we fail with g++-4.*.

$ docker run --rm -ti rocker/r-apt:precise g++ --version | head -1         # bad
g++ (Ubuntu/Linaro 4.6.3-1ubuntu5) 4.6.3
$ docker run --rm -ti rocker/r-apt:trusty g++ --version | head -1          # bad
g++ (Ubuntu 4.8.4-2ubuntu1~14.04.3) 4.8.4
$ docker run --rm -ti rocker/r-apt:wily g++ --version | head -1            # good
g++ (Ubuntu 5.2.1-22ubuntu2) 5.2.1 20151010
$ docker run --rm -ti rocker/r-apt:xenial g++ --version | head -1          # good
g++ (Ubuntu 5.4.0-6ubuntu1~16.04.4) 5.4.0 20160609
$ 

Maybe that version is common enough to warrant a look.

(If I may advertise a little: that Docker container series 'rocker-apt' is tremendously useful, but under-advertised and even I used them less than I should. To be changed...)

[krlmlr]

Wasn't aware of rocker-apt. Thanks for looking into this.

I'm afraid this might affect many packages that use Rcpp and test on Travis.

[eddelbuettel]

Well only if they test of for equality of an expected message. ;-)

But as I demonstrated outside of Travis, it really has nothing to do with Travis. But maybe with the g++ version and interactions with tinyformat or whatnot.

[jimhester]

@lionel- ran into this as well https://travis-ci.org/lionel-/svglite/builds/177153874#L2484, I failed to reproduce it but I was using an image later than precise, so good to know we have an example.

[eddelbuettel]

I guess it also means that we don't have a test checking for equality of the returned text, or else we would have know from testing on 'trusty' (aka 14.04).

[jimhester]

We check for the message inst/unitTests/runit.exceptions.R#L33 but all of the checked exceptions are either standard c++ exceptions or an explicit Rcpp exception. This issue seems to only occur with exceptions generated from Rcpp::stop(), which is not tested :(.

I am looking into fixing this now.