[PATCH v2] edit: Return error when save_file failed

[Antony Pavlov]

On Wed, 4 Nov 2015 14:30:53 +0100
Sascha Hauer <s.hauer at pengutronix.de> wrote:

> On Wed, Nov 04, 2015 at 03:23:21PM +0300, Antony Pavlov wrote:
> > On Wed,  4 Nov 2015 12:03:24 +0100
> > Enrico Jorns <ejo at pengutronix.de> wrote:
> > 
> > > When writing a file failed (e.g. due to a read-only file system), no
> > > error was reported by the 'edit' tool. To be valid (and to not confuse
> > > the poor user) at least '1' should be returned to indicate an error.
> > > 
> > > Signed-off-by: Enrico Jorns <ejo at pengutronix.de>
> > > ---
> > >  commands/edit.c | 5 +++--
> > >  1 file changed, 3 insertions(+), 2 deletions(-)
> > > 
> > > diff --git a/commands/edit.c b/commands/edit.c
> > > index b28e2b9..3b653b7 100644
> > > --- a/commands/edit.c
> > > +++ b/commands/edit.c
> > > @@ -375,6 +375,7 @@ static int do_edit(int argc, char *argv[])
> > >  	int i;
> > >  	int linepos;
> > >  	int c;
> > > +	int ret = 0;
> > 
> > Can we use COMMAND_SUCCESS insted of 0?
> > 
> > 
> > >  
> > >  	if (argc != 2)
> > >  		return COMMAND_ERROR_USAGE;
> > > @@ -533,7 +534,7 @@ static int do_edit(int argc, char *argv[])
> > >  			}
> > >  			break;
> > >  		case 4:
> > > -			save_file(argv[1]);
> > > +			ret = save_file(argv[1]);
> > 
> > Actually save_file() returns open()'s error. We have to convert it to COMMAND_ERROR, e.g.:
> 
> If we return an error code from a command then the caller will print the
> corresponding error string to the console which might be what we want
> here.
> 
> > 
> > if (save_file(argv[1]) != 0)
> > 	ret = COMMAND_ERROR;
> > 

At the moment the only cmd() caller is inside of execute_command():

        ret = cmdtp->cmd(argc, argv);
        if (ret == COMMAND_ERROR_USAGE) {
                barebox_cmd_usage(cmdtp);
                ret = COMMAND_ERROR;
        }
        ...
        return ret;


But we brop execute_command() result in run_command():
        if (execute_command(argc, argv) != COMMAND_SUCCESS)
	        rc = -1;

We have a chance to see error message only if we call execute_command()
from run_pipe_real():

        ret = execute_binfmt(globbuf.gl_pathc, globbuf.gl_pathv);
        if (ret < 0) {
            printf("%s: %s\n", globbuf.gl_pathv[0], strerror(-ret));
            ret = 127;
        }

So, yes, you are right we have no reason to convert open()'s error into COMMAND_ERROR,
but for consistency we have toadd strerror() print to run_command() code.

> > Also save_file() does not check write()'s error.
> Indeed, this could be fixed while touching this code.
> 
> Sascha
> 
> 
> -- 
> Pengutronix e.K.                           |                             |
> Industrial Linux Solutions                 | http://www.pengutronix.de/  |
> Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0    |
> Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |


-- 
-- 
Best regards,
  Antony Pavlov