>>>>> Peter Dalgaard writes:
> Kurt Hornik <[EMAIL PROTECTED]> writes:
>> > (Argh. The test builds of 1.8.0 were actually against tcl8.0, but
>> > apparently our testing procedures skip any attempt to load the dynlib.
>> > Automated tests of tcl/tk are difficult because Tk will protest if
>> > there is no X display.)
>>
>> On systems which have Xvfb, this is not a problem.
>>
>> The daily package checker uses this, for example.
> Hmm it is more tricky than that it seems. I see
> comparing 'stepfun-Ex.Rout' to 'stepfun-Ex.Rout.prev' ... OK
> running code in 'tcltk-Ex.R' ... OK
> comparing 'tcltk-Ex.Rout' to 'tcltk-Ex.Rout.prev' ... OK
> collecting examples for package 'ts' ...
> make[5]: Entering directory `/e7/pd/r-devel/BUILD/src/library'
> However,
> [EMAIL PROTECTED] BUILD]$ tail tests/Examples/tcltk-Ex.Rout
>> postscript("tcltk-Examples.ps")
>> assign("par.postscript", par(no.readonly = TRUE), env = .CheckExEnv)
>> options(contrasts = c(unordered = "contr.treatment", ordered = "contr.poly"))
>> require('tcltk') || q()
> Loading required package: tcltk
> Error in dyn.load(x, as.logical(local), as.logical(now)) :
> unable to load shared library
> "/e7/pd/r-devel/BUILD/library/tcltk/libs/tcltk.so":
> /e7/pd/r-devel/BUILD/library/tcltk/libs/tcltk.so: undefined symbol:
> RTcl_ActivateConsole
> Error in runHook(".onLoad", env, package.lib, package) :
> .onLoad failed
> I.e. the seeming pass of make check is really just reflecting the fact
> that I got the same catastrophic failure as last time I ran it...
> Is there a better way?
[Untested.]
Maybe we want to try to look more closely at why require("tcltk") might
have failed, as opposed to always exiting gracefully on failure via
require('tcltk') || q()
Something like
yy <- try(require("tcltk"), silent = TRUE)
if(any(regexpr("^Tcl/Tk support is not available") > -1))
q()
else
stop(yy)
???
-k
______________________________________________
[EMAIL PROTECTED] mailing list
https://www.stat.math.ethz.ch/mailman/listinfo/r-devel
