Re: CVS commit: src/sbin/gpt

[christos%astron.com@localhost (Christos Zoulas)]

In article <201512010058.tB10wTZJ029677%server.CornerstoneService.ca@localhost>,
John Nemeth  <jnemeth%cue.bc.ca@localhost> wrote:
>     See question in gpt.c (and others).
>
>} -	printf("Partition %d added, use:\n", i + 1);
>} -	printf("\tdkctl %s addwedge <wedgename> %" PRIu64 " %" PRIu64
>} -	    " <type>\n", device_arg, map->map_start, map->map_size);
>} -	printf("to create a wedge for it\n");
>} +	printf("Partition %d added on %s: ", i + 1, device_arg);
>} +	printf("%s %" PRIu64 " %" PRIu64 "\n", type, map->map_start,
>} +	    map->map_size);
>
>     This message needs to depend on the new -n flag, otherwise it
>may give incorrect information.

I will add a message to clarify that thanks!

>} @@ -269,10 +271,11 @@ gpt_write(int fd, map_t *map)
>}  
>}  	count = map->map_size * secsz;
>}  	ofs = map->map_start * secsz;
>} -	if (lseek(fd, ofs, SEEK_SET) == ofs &&
>} -	    write(fd, map->map_data, count) == (ssize_t)count)
>} -		return (0);
>} -	return (-1);
>} +	if (lseek(fd, ofs, SEEK_SET) != ofs ||
>
>     Why did you change the "&&" to "||"?  If the lseek() fails,
>you will now be writing data to the wrong place on the disk.

Nope, I flipped the polarity of the tests too.

>
>} +	    write(fd, map->map_data, count) != (ssize_t)count)
>} +		return -1;
>} +	modified = 1;
>} +	return 0;

christos