From: lars@beagle-ears.com   
      
   On 2026-01-07, Carlos E.R.  wrote:   
   > [Pascal strings] those store the size of the used string as the first   
   > byte. We also need the maximum size of the string variable.   
   > The compiler knows.   
   >   
   > Turbo Pascal generated code to check the bounds of a variable in runtime   
   > and aborted the program if surpassed. I was amazed at seeing production   
   > code disabling the check (for speed?).   
      
   Of course, it is possible to bounds check every access. Incurring a huge   
   loss of runtime performance. And maybe that should be a compiler option.   
      
   In most programs of any size, most strings are dynamically allocated,   
   so the compiler does NOT know their size. The runtime COULD know it from   
   the malloc() header, but then they are different from static strings.   
   You COULD paper that over by having the header on static strings, but   
   then you lose the ability to have a string in a packed struct.   
      
   C++ eliminates those issues by allowing you to define a string class   
   that must be allocated by "new", but string handling is so widely used   
   that you'd have to rewrite a LOT of runtime libraries.   
      
   Much of the performance power of C comes from NOT incurring the overhead   
   of this safety net.   
      
   ** digression **   
      
   In my own world, the real-time control of radio devices with   
   microsecond packet timing prescheduled so we can account for   
   time-of-flight (speed of light x distance) depends on knowing   
   the maximum interrupt latency and being able to limit the time   
   interrupts are disabled. We count instructions in the interrupt   
   service routines from the time the serial framer says "packet to   
   my address detected" to the point where we can set the trigger   
   that allows the "end-of-packet" to un-gate the transmit clock.   
      
   When Motorola got out of the chip business, first spinning it   
   out as FreeScale, then passing it off to NXP, the support for   
   bare-bones developers disappeared for new members of the family,   
   and they more or less insisted that the embedded products had   
   to run Linux.   
      
   We decided not to move to the newer versions. For over a decade,   
   we have been able to live off the last-time-buy   
   of each chip that went out of production.  The approaching   
   end of that stock fortuitously dovetails with our desire to   
   retire and shut down our business - with proper time buffer   
   to allow us to give our customers 2 years of warning and   
   make their own last-time-buy. But it is hard to convince a railway   
   company in Taiwan that we REALLY are going away.   
   --   
   Lars Poulsen - an old geek in Santa Barbara, California   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)