FreeTDS Development Group

["ZIGLIO, Frediano, VF-IT" <Frediano.Ziglio AT vodafone.com>]

> 
> Hi there,
> 
> thanks for freetds!
> 
> I use it since 2 years and today I ran into a problem with 
> 0.63 (from the
> debian package - I ran 0.61 before and it had the same problem):
> 
> I'm reading a record with a datetime and I don't know why but 
> a special row
> seems to contain garbage at the end:
> 
> src=0xbfffe030 "Di 10 Mai 2005 14:03:43 š%S\bhàÿ¿ËÍî·8"
> (german locale)
> 

If src point to row buffer this can be normal, strings in row buffer are not 
null terminated.

> Then the parsing routine set the timezone to an illegal value:
> 
> tm = {tm_sec = 43, tm_min = 3, tm_hour = 14, tm_mday = 10, tm_mon = 4,
>   tm_year = 105, tm_wday = 2, tm_yday = 130, tm_isdst = 0, 
> tm_gmtoff = 420,
>   tm_zone = 0x2b <Address 0x2b out of bounds>}
> 
> and the program crashed with a segmentation fault. What I did 
> is to set the
> timezone explicitly. I did a call to tzset() and set 
> tm.tm_zone = tzname[0]
> in convert.c in tds_strftime().
> 

You are right

in tds_strftime

size_t
tds_strftime(char *buf, size_t maxsize, const char *format, const TDSDATEREC 
* dr)
{
        struct tm tm;

        int length = 0;
        char *s, *our_format;
        char millibuf[8];

        char *pz = NULL;

        tm.tm_sec = dr->second;
        tm.tm_min = dr->minute;
        tm.tm_hour = dr->hour;
        tm.tm_mday = dr->day;
        tm.tm_mon = dr->month;
        tm.tm_year = dr->year - 1900;
        tm.tm_wday = dr->weekday;
        tm.tm_yday = dr->dayofyear;
        tm.tm_isdst = 0;

so we do not reset tm.tm_zone ... the solution is to add a 

        tm.tm_zone = NULL;

line (I never though strftime could use tm_zone field).
I'll add a fix to 0.63 (and 0.64)

> I don't know if the problem is somewhere else such that this 
> crazy timestamp
> gets through at all but tm_zone did not get initialized and 
> it should be...
> It seems to be a gnu extension however (I'm running debian 
> sarge on i386).
> 
> I attach a more detailed gdb output below.
> 
> A TDSDUMP-log (from another example however) is attached as well.
> 
> I have another question: Is there any way I can set a 
> different datestyle
> than via locale? The problem is that if you follow your locale the sql
> server might not recognize this input (i.e. its own output) 
> as a valid date
> format which is pretty annoying... What do you think about 
> this problem?
> 
> 
> Joachim
> 

bye
  freddy77