Wireshark · Ethereal-dev: Re: [Ethereal-dev] Not possible to capture packets on win32 on latest svn build?

[Ulf Lamping <ulf.lamping@xxxxxx>]

Guy Harris wrote:

> Not exiting before *re*-scanning the command line parameters is *also* a
> bad idea, as one error message is better than two, and getopt() will print
> error messages for at least some command-line problems.
>  
Hmmm, so both ways don't work well :-(

> (Using g_warning() for command-line errors is also a bit of a bad idea, as
> it makes the error messages *REALLY* ugly:
>
>    $ ethereal -tq
>    14:19:21          Warn ethereal: Invalid time stamp type "q"
>    14:19:21          Warn It must be "r" for relative, "a" for absolute,
>    14:19:21          Warn "ad" for absolute with date, or "d" for delta.
>
> The time stamp and "Warn" serve no purpose and just clutter the output,
> making it harder to read.)
>  
That's ugly, right. But having *no* output at all on Win32 is even much 
worse. That's the reason I've changed it that way, it was simply a quick 
fix.

> If there's a case where getopt() returns '?' in the capture child, that's
> an error.  If the command-line arguments are different, the option string
> should be different, too.
>  
The option strings are different, it's changed for the second check:

 /* Set "capture_child" to indicate whether this is going to be a child
    process for a "-S" capture. */
 capture_child = (strcmp(command_name, CHILD_NAME) == 0);
 if (capture_child) {
   strcat(optstring, OPTSTRING_CHILD);
 }

We might need to move this way up before the first check. I can't 
remember if there was a reason not to do so.

> The fix is to print a usage message in that case.  I'm compiling that fix
> now, although I have to take off for a medical appointment and won't be
> back for a while, so I'll check it in later.
>  
Shouldn't be too difficult.

Regards, ULFL