andrew clarke wrote in a message to Roy J. Tellason:

 ac> Sat 2003-06-21 12:08, Roy J. Tellason (1:270/615) wrote to andrew
 ac> clarke: 

 RJT> There are a couple of bits of that program you posted that I'm not
 RJT> sure of.  For example,  this stuff:

 RJT> #ifdef __GNUC__
 RJT> #include 
 RJT> #include 
 RJT> #else
 RJT> #include 
 RJT> #include 
 RJT> #endif

 ac> If compiling under GNU C, include unistd.h & utime.h, otherwise
 ac> include io.h and sys/utime.h.  Different brands of compilers can be
 ac> inconsistent as to where certain prototypes and other things (that
 ac> are not in the ISO C standard) are located.

I got the conditional,  I just don't know what those headers are for.  I
suppose I could find them and have a look?  :-)

 RJT> static struct stat s_buf;

 ac> Declare a variable of type struct stat and call it s_buf.  Make it
 ac> static, ie. local to this .c file only, which has the side effect
 ac> of setting all its values to zero during program startup.

That side effect is an interesting tidbit,  I hadn't remembered that.  But
then I wouldn't have expected any initialization to take place unless I put
some in there,  so not remembering that wouldn't have been a problem.

 RJT> and these bits here:

 RJT> ut_buf.actime = s_buf.st_atime;
 RJT> ut_buf.modtime = s_buf.st_mtime;

 RJT> What are those for?

 ac> s_buf is obtained from the file system by using stat().  Among 
 ac> other things struct stat contains the file's "last access" time and
 ac> "last modified" time.  struct utimbuf is used for writing this 
 ac> information back to the file system by way of the utime() function.

Ok,  there are a couple of things I wasn't familiar with...

 ac> So we're setting the timestamp of the newly create file to the 
 ac> timestamp of the original file.

I like it!

 ac> Then we delete the old file and rename the new file to the old 
 ac> file's name.

Yep.  The thing as a whole works well,  and I can see where the basic
framework might be useful for some other things as well.

---