Mail Archives: cygwin/2003/01/01/18:56:09
Christopher Faylor wrote:
> On Wed, Jan 01, 2003 at 12:19:52PM -0500, Pierre A. Humblet wrote:
>
>>>On Wed, Jan 01, 2003 at 10:05:10AM +0000, Steven O'Brien wrote:
>>
>>>>In glib-1.2.10, gutils.c: g_get_any_init (void), the current user
>>>>details are obtained from /etc/passwd. This code is called as part of
>>>>glib initialisation, whether the app wants this data or not. It uses
>>>>sysconf (_SC_GETPW_R_SIZE_MAX) to decide how much buffer to allocate for
>>>>this data. But on cygwin this appears broken, and the call to getpwuid_r
>>>>(getuid (), &pwd, buffer, bufsize, &pw) *may* overrun buffer, depending
>>>>on the length of the line in /etc/passwd for the current user.
>>>
>>Have you experienced an overrun? I see code in cygwin to set errno to
>>ERANGE if the buffer is too small.
>
>
> Aha! There's the bug. AFAICT, the size calculation is wrong. It isn't
> taking the size of all of the elements into consideration.
No comment on this stuff. It's not the problem I am seeing.
The _SC_GETPW_R_SIZE_MAX thing was introduced after glib-1.2.8. In
1.2.8 the buffer size was just 64. In later versions, glib tries the
sysconf() call if available, otherwise uses some other set of defaults
for various systems.
In any case, 64 is too small for cygwin -- but mebbe if there IS a bug
in sysconf ( _SC_GETPW_R_SIZE_MAX ), then cgf's changes to cygwin will
fix that.
But my bug is something else -- and it may be in newlib, but I suspect
glib's g_printf_string_upper_bound function (and calling vsprintf
instead of vsnprintf)
--Chuck
--
Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple
Bug reporting: http://cygwin.com/bugs.html
Documentation: http://cygwin.com/docs.html
FAQ: http://cygwin.com/faq/
- Raw text -