Re: 01-getnline.diff (was Re: getline & getline_safe)

On Wed, Jul 16, 2003 at 03:56:33PM -0400, Derek Robert Price wrote:
>
>Okay, here's my first pass.  It configures, builds, and passes `make 
>check' with CVS, so I updated modules/getline, created modules/getnline, 
>and am resubmitting to bug-gnulib.  Please let me know what you think.
>
>I call it a first pass only because I wasn't sure how to document this 
>complex dependency.  I think what I did was acceptable, but notice that 
>I duplicated the getnline.* stuff between the two module descriptions.  
>The basic difference is that if you want getline compiled conditionally 
>but getnline compiled unconditionally then you add a call to 
>gl_PREREQ_GETNLINE to configure.ac.  This is what I did with CVS.  If 
>you just include the info as per the modules/getline file, getnline.c 
>will only be compiled when it is needed for getline.c.
>
>I still need to write the ChangeLog for CVS but I should commit these 
>changes there soon.
>
>Incidentally, my glibc defines both getdelim & getline to return 
>ssize_t.  Is it intentional that the GNULIB versions return int instead?

Actually, that was exactly the cause of the original bug! ssize_t and
int are equivalent on most 32-bit platforms, but not on 64-bit
Linux. Thus the definition of getline() clashes and does not match the
one included from stdio.h...

-- 
Steve McIntyre, Cambridge, UK.                                [email protected]
"It's actually quite entertaining to watch ag129 prop his foot up on
 the desk so he can get a better aim."          [ seen in ucam.chat ]