R devel - Aug 2003 - round.POSIXt sometimes crashes R (PR#3763)

Full_Name: andrea  capodicasa
Version: 1.7.0
OS: w2k sp3
Submission from: (NULL) (212.17.194.154)


Hi all,
the problem is when you try to round.Posix an empty vector of dates
  
Please give a look to this code:
> data=seq(ISOdate(2001,1,1),by="day",length=3)
> data
[1] "2001-01-01 13:00:00 ora solare Europa occidentale"
[2] "2001-01-02 13:00:00 ora solare Europa occidentale"
[3] "2001-01-03 13:00:00 ora solare Europa occidentale"

this works:
> non.crasha=round(data[c(F,F,T)],"day")
> non.crasha
[1] "2001-01-04"


this (subsetting data with an empty set) crashes R on w2k
crasha=round(data[c(F,F,F)],"day")

these are infos about my R
> R.version
         _              
platform i386-pc-mingw32
arch     i386           
os       mingw32        
system   i386, mingw32  
status                  
major    1              
minor    7.0            
year     2003           
month    04             
day      16             
language R       

---
TIA
Andrea Capodicasa

Grazie mille, Andrea,
for the bug report!
>>>>> "AndreaC" == andrea capodicasa
<andrea.capodicasa@albacom.it>
>>>>>     on Mon, 18 Aug 2003 16:21:45 +0200 (MET DST) writes:
    AndreaC> Full_Name: andrea capodicasa Version: 1.7.0 OS: w2k
    AndreaC> sp3 Submission from: (NULL) (212.17.194.154)


    AndreaC> Hi all, the problem is when you try to round.Posix
    AndreaC> an empty vector of dates
  
I can confirm the problem on Linux, using a much newer version
of R (R-devel of today).

A bit shorter version:
> round(ISOdate(2001,1,1)[0], "day")
Process R gleitkomma-Ausnahme at Mon Aug 18 17:23:44 2003
	  ^^^^^^^^^^^^^^^^^^^
( = locale-translation of  "floating point exception" )

--------------
The same for Linux and Solaris:
Using  "R -d gdb" and "bt" (= backtrace) after the error
reveals

(gdb) bt
#0  0xfefbee9c in .rem () from /usr/lib/libc.so.1
#1  0x0004e4b4 in do_formatPOSIXlt (call=0xebc9c8, op=0xebc9c8, args=0xed2fb8, 
    env=0x1) at ../../../R/src/main/datetime.c:565
......
.......

-----------------

There's an obvious way to fix this bug on the R- (as opposed to C-)
level, and I'd probably just do it.
However, I wonder if the logic in the above C code is okay..

Martin Maechler <maechler@stat.math.ethz.ch>
http://stat.ethz.ch/~maechler/
Seminar fuer Statistik, ETH-Zentrum  LEO C16	Leonhardstr. 27
ETH (Federal Inst. Technology)	8092 Zurich	SWITZERLAND
phone: x-41-1-632-3408		fax: ...-1228			<><