Dear all, Brian Ripley has brought it my attention that R 3.1.2, due to be released soon, *does* check that X11 is present and prints an informative error message if it isn't. That's a substantial improvement.
Thanks, John On Tue, 28 Oct 2014 09:43:53 -0400 "John Fox" <[email protected]> wrote: > Dear Rich, > > I've made a similar suggestions before, I believe. It would even help to have > tcltk fail less cryptically if X-Windows is unavailable under Mac OS X, > suggesting that XQuartz be installed and pointing to the XQuartz website. And > it shouldn't be hard to check capabilities()["X11"] when tcltk loads. > > The large majority of Rcmdr problems about which people write me concern this > one issue. There are clear, detailed installation instructions on the Rcmdr > website but no way for me to insure that people are aware of them. I spent > some time trying to find a way to report the absence of X-Windows at Rcmdr > start-up, but the dependency on tcltk causes the Rcmdr to fail to load before > an informative message can be printed -- a chicken-and-egg problem. Maybe > there's a way to do this that I didn't discover. > > The essential problem here is that naive users don't have an obvious way of > avoiding the problem. They install the software and expect it to work. There > are similar issues with the system path under Yosemite and with app nap. This > leads to repetitious emails to package maintainers and to repetitive > questions on the various help lists, etc. > > That said, I think that there is a good solution for teachers of courses, > workshops, etc.: Create -- or point to existing -- clear installation > instructions that include installing XQuartz under OS X. The less tractable > problem is individual users downloading and installing R from CRAN. > > Best, > John > > On Tue, 28 Oct 2014 00:03:52 -0400 > "Richard M. Heiberger" <[email protected]> wrote: > > I just did a workshop for new R users. the only serious installation > > problem I > > saw was Macintosh users and tctlk. From this discussion, I knew to tell > > both > > people they needed to download quartz.macosforge.org > > and indeed that solved it. > > > > Is there a place on the package DESCRIPTION file to state that dependency, > > thus triggering the download and installation of quartz? That would > > take the responsibility for discovering this need away from the end user? > > > > Rich > > > > > > On Thu, Oct 23, 2014 at 2:15 PM, John Fox <[email protected]> wrote: > > > Dear Simon, > > > > > > I installed Yosemite a couple of days ago and everything seems to work > > > fine so far, including the tcltk demo that caused problems for Peter, and > > > the Rcmdr package, which gives Tcl/Tk a pretty good workout. I first > > > reinstalled XQuartz, as suggested, and I also reinstalled R and updated > > > all packages, though the latter two steps probably weren't necessary. I > > > figured that it would help to hear positive experiences as well as > > > problems. > > > > > > The only issue that I've encountered so far is specific to checking > > > packages under RStudio, which doesn't appear to find pdflatex; OTOH, R > > > CMD check runs fine in a terminal window. I haven't yet tried to resolve > > > this problem. > > > > > > Best, > > > John > > > > > >> -----Original Message----- > > >> From: Simon Urbanek [mailto:[email protected]] > > >> Sent: Tuesday, October 21, 2014 4:38 PM > > >> To: John Fox > > >> Cc: Amos B. Elberg; r-sig-mac; Spencer Mass; Ben Clark; peter dalgaard; > > >> Marc Schwartz; David Winsemius; Hadley Wickham > > >> Subject: Re: [R-SIG-Mac] Yosemite and R > > >> > > >> I wasn't able to reproduce but I suspect those are all red herrings - > > >> there are really no subprocesses involved at all in either case. > > >> > > >> On Oct 21, 2014, at 4:19 PM, John Fox <[email protected]> wrote: > > >> > > > >> > Dear all, > > >> > > > >> > I wonder whether this issue also accounts for the tcltk problems that > > >> have been reported. (I haven't yet upgraded to Yosemite myself, hoping > > >> to wait for the wrinkles to be ironed out, though I'll likely do so > > >> shortly if only to see what happens.) > > >> > > > >> > Best, > > >> > John > > >> > > > >> >> -----Original Message----- > > >> >> From: [email protected] [mailto:r-sig-mac-bounces@r- > > >> >> project.org] On Behalf Of Amos B. Elberg > > >> >> Sent: Tuesday, October 21, 2014 12:40 PM > > >> >> To: David Winsemius; Hadley Wickham > > >> >> Cc: r-sig-mac; Spencer Mass > > >> >> Subject: Re: [R-SIG-Mac] Yosemite and R > > >> >> > > >> >> If the full environment isnt getting passed to R-spawned sub- > > >> >> processes, that might explain an error Ive been having since the > > >> >> update: when R is launched from the command line, calls that should > > >> >> create an X11 window in the background fail unless an X11 window has > > >> >> already been created with the width and height specified: > > >> >> > > >> >>> plot(rnorm(100)) > > >> >> Error in .External2(C_X11, d$display, d$width, d$height, > > >> >> d$pointsize, : > > >> >> invalid 'width' or 'height' > > >> >>> X11() > > >> >> Error in .External2(C_X11, d$display, d$width, d$height, > > >> >> d$pointsize, : > > >> >> invalid 'width' or 'height' > > >> >>> X11(width = 5, height = 5) > > >> >>> plot(rnorm(100)) > > >> >> [now it works - and any number of additional windows can be spawned > > >> >> without repeating the error] > > >> >> [quitting R and reopening, without quitting XQuartz, and I get the > > >> same > > >> >> error calling plot() before X11(width = , height = ) > > >> >> > > >> >> This does not happen in RStudio. I dont use the R.app gui; I > > >> opened > > >> >> it just now to test and I got a slew of path-related errors, but > > >> >> theyre as likely to have to do with my not-maintained R.app as with > > >> >> anything else. > > >> >> > > >> >> I had not reported this already because I wasnt confident whether > > >> its > > >> >> a yosemite issue, an R-patched issue, or just something odd in the > > >> way > > >> >> I built R. > > >> >> > > >> >> If incomplete-environment-passing is the new normal, is this not > > >> going > > >> >> to be a common issue for packages that spawn sub-processes? > > >> >> > > >> >> > > >> >> From: Hadley Wickham <[email protected]> > > >> >> Reply: Hadley Wickham <[email protected]>> > > >> >> Date: October 21, 2014 at 10:12:13 AM > > >> >> To: David Winsemius <[email protected]>> > > >> >> Cc: r-sig-mac <[email protected]>>, Spencer Mass > > >> >> <[email protected]>> > > >> >> Subject: Re: [R-SIG-Mac] Yosemite and R > > >> >> > > >> >>> No, it is not. It is expected that the path in the terminal be > > >> >>> different to the path in R, it is _not_ expected that the path in R > > >> >> be > > >> >>> different to the path in a subprocess started by R. > > >> >>> > > >> >>> (Well it is now expected, because this appears to be a new security > > >> >>> feature in Yosemite) > > >> >> > > >> >> The best thread I could find on the problem is here: > > >> >> https://code.google.com/p/mactlmgr/issues/detail?id=102 > > >> >> > > >> >> -- > > >> >> http://had.co.nz/ > > >> >> > > >> >> _______________________________________________ > > >> >> R-SIG-Mac mailing list > > >> >> [email protected] > > >> >> https://stat.ethz.ch/mailman/listinfo/r-sig-mac _______________________________________________ R-SIG-Mac mailing list [email protected] https://stat.ethz.ch/mailman/listinfo/r-sig-mac
