Re: question about linker

On 07/12/2024 14:51, Tim Rentsch wrote:

Bart <bc@freeuk.com> writes:
 

In a 39Kloc C codebase, there were 21,500 semicolons.
>
This is generated C, so some of those may be following labels!
2000, actually, which still leaves 19,500; a lot of semicolons.
>
On a preprocessed sql.c test (to remove comments) there were
53,000 semicolons in 85,000 lines.
>
So in C, they are a very big deal, occuring on every other line.

 Many years ago I read a book called "How to Lie with Statistics".
It should be updated to add this reporting as an example.

You're welcome to post your own figures.
I know your own code will be out of the ordinary as you prefer using commas (or ?:) to semicolons, and to write things horizontally rather than vertically.
Here's one of my few programs that I manually wrote in C:
   c:\cx>mcc pid
   Compiling pid.c to pid.exe
   NLINES= 2219
   NSEMIS= 1078
   NSEMISATEOL= 1026
It is pretty much exactly what I said. One more (not mine):
   c:\cx>mcc deltablue
   Compiling deltablue.c to deltablue.exe
   NLINES= 1558
   NSEMIS= 707
   NSEMISATEOL= 652
This one is a bit different:
   c:\cx>mcc piet
   Compiling piet.c to piet.exe
   NLINES= 3267
   NSEMIS= 753
   NSEMISATEOL= 680
Semicolons occur only every 4 lines or so. Hmm, let's try preprocessing, then compile the processed source:
   c:\cx>mcc -e piet
   Preprocessing piet.c to piet.i
   c:\cx>mcc piet.i
   Compiling piet.i to piet.exe
   NLINES= 1331
   NSEMIS= 774
   NSEMISATEOL= 719
Back to normal! I guess there were a lot of #defines, or something. It's difficult to isolate large, external headers from line counts. But that's not code that /you/ have to write; yours will be the executable code that needs the semicolons!