[Rd] Re: Sys.time dumps core at start-up (PR#822)

Peter Kleiweg kleiweg@let.rug.nl
Wed, 24 Jan 2001 10:30:31 +0100 (CET)

Prof Brian Ripley skriver...

> On Tue, 23 Jan 2001, Peter Kleiweg wrote:
> > Prof Brian D Ripley skriver...

> > > Did you set TZ as advised in ?DateTimeClasses to work around
> > > bugs in deficient Linux distros? RH6.2's strftime will crash
> > > on POSIXly (and its man page) correct inputs.
> >
> > Setting TZ has no effect. R still dumps core. I don't see the
> > point of setting TZ, since it is unset in datetime.c
> It is only unset (if possible) if it was initially unset.

There might be some logic behind this, but don't expect every
outsider to understand this. I just noticed setting TZ has no

> You do seem to
> give the idea that you think the R developers do not know what they are
> doing! Did you actually test your hypothesis?  If you did, and it is
> correct, that is another bug of your system. On RH6.2

I read the code in datetime.c. In there TZ is unset. I have no
overview on how this code is used.
> > In another mail you ask me to debug the problem. I have no
> > experience debugging large programs, and if the problem is
> > caused by a faulty library function I don't see the point
> > trying.
> But you did send a debug log to R-devel shortly after I asked
> which told me the error was in strftime....  It would have been
> helpful to have put that (and your Linux distro) in the original message.

I only copied something I saw being done in another mail I
received just after I sent my first message. As to debugging the
complete code, see how it got to this error, that is something
beyond me.
> Please do remember that there is not a technical support section receiving
> these bug reports, but a group of volunteers,  those who in the main
> provided you with the software in the first place.  Incomplete and
> inaccurate reports make their task harder than necessary, and that's
> unkind.

Well, I'm not a professional either. I don't file bug reports
demanding a solution. I thought bug reports might be useful to
the developers. Perhaps you should add a warning to the help on
bug.report, something like "don't bother submitting a bug report
if you haven't a clue".

Peter Kleiweg

r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-devel-request@stat.math.ethz.ch