octave 2.9.13 on cygwin

[John W. Eaton]

On 30-Aug-2007, Marco Atzeri wrote:

| Hi,
| first I would like to confirm that 
| octave-2.9.13 compile fine on latest 
| cygwin snapshot 1.7.0s(0.177/4/2) 20070813 12:16:31.
| 
| Of course cs.h patch of Tatsuro is needed.
| A minor issue due to cygwin snapshot headers
| http://cygwin.com/ml/cygwin/2007-08/msg00740.html
| should be corrected by cygwin developers.
| 
| However both my build, both Tatsuro are failing
| check tests related to date/calendar. 
| 
| I presume it is due to function strptime not working 
| as expected, at least on my cygwin.
| Can anyone confirm it for others cygwin ?
| 
| [tm,nc]=strptime('30-Aug-2007','%d-%b-%Y')
| tm =
| {
|   usec = 0
|   sec = 0
|   min = 0
|   hour = 0
|   mday =  30
|   mon =  7
|   year =  107
|   wday = 0
|   yday = 0
|   isdst = 0
|   zone = WEST
| }
| 
| nc =  12
| 
| It seems a old issue not solved yet
| http://www.cae.wisc.edu/pipermail/bug-octave/2007-March/001821.html

In that message, Paul Probert wrote:

| I looked at datevec.m and did a little debugging. It seems the error is 
| in strptime.oct. If you do:
| 
| octave:11> [tm,nc]=strptime('03-Mar-2007','%d-%b-%Y')
| tm =
| {
|    hour = 0
|    isdst = 0
|    mday = 3
|    min = 0
|    mon = 2
|    sec = 0
|    usec = 0
|    wday = 6
|    yday = 61
|    year = 107
|    zone =
| }
| 
| nc = -11
| 
| 
| Note that it works fine but the nc is negative, whereas datevec is 
| expecting the length of the input string plus 1, ie 11.
| I get this error on strptime in 2.1.73 on linux also.
| 
| Now I'm lost when trying to find where strptime.oct comes from, but 
| liboctave/oct-time.cc I found a line (350) where
| nchars = p - q;
| 
| and it seems to me this should be q - p;
| 
| This doesn't touch on my original cygwin 2.1.73 error with datevec, but 
| maybe some more knowlegeable people might want to look at strptime in 
| your current version. I didn't check to see if this is fixed in the 2.9 
| series.

The current code in the function octave_strptime::init in
liboctave/oct-time.cc is:

  char *p = strsave (str.c_str ());

  char *q = oct_strptime (p, fmt.c_str (), &t);

  if (q)
    nchars = q - p + 1;
  else
    nchars = 0;

  delete [] p;


Can you please debug why it fails to compute the correct number of
characters on your system?

jwe