From: arne@vajhoej.dk   
      
   On 6/23/2025 9:41 PM, Lawrence D'Oliveiro wrote:   
   > On Mon, 23 Jun 2025 20:40:21 -0400, Arne Vajhøj wrote:   
   >> On 6/23/2025 8:37 PM, Lawrence D'Oliveiro wrote:   
   >>> What a dumb way to do it. Using the null device means making OS I/O   
   >>> calls -- needless overhead just to find out how long a string -- which   
   >>> is being put together entirely in userland -- needs to be.   
   >>   
   >> But if you have a simple better way of achieving it ...   
   >   
   > snprintf(3) and vsnprintf(3) return the total number of bytes needed for   
   > the final result. So just call them once with all the right arguments and   
   > a zero-length buffer, and use the result to allocate a buffer of the right   
   > length, and call them again.   
      
   That is a lot better than vfprintf to null.   
      
   But it does not solve the problem of multiple v*printf calls.   
      
   > And of course you should be using va_end(3) and va_start(3) again before   
   > the second call. This allows for traversing the argument list multiple   
   > times.   
      
   Yes. But the argument list for vasprintf is as it is. This is   
   code replacing a missing RTL function with a given argument list.   
      
   So calling va_* multiple times means switching from a replacement   
   function to #ifdef code in calling function (dbfcmd).   
      
   Arne   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)