[U-Boot] 'run' no longer aborts on an error, can someone please verify

[Timo Ketola]

Dear Gray, Jason, Heiko, Simon,

I added you, Jason, Heiko and Simon, into the CC-list because there is
your commits in the history which seems somewhere to change the
behaviour below:

On 22.04.2012 00:55, Gray Remlin wrote:
> U-Boot v2012.04
> 
> Marvell>> setenv ipaddr '192.168.1.130'
> Marvell>> setenv ifhostisup 'ping 192.168.1.100'
> Marvell>> setenv saywearehappy 'echo "We are happy!"'
> Marvell>> run ifhostisup saywearehappy;
> Using egiga0 device
> ping failed; host 192.168.1.100 is not alive
> "We are happy!"
> Marvell>> 
> 
> This is not the same behaviour as my (unfortunately customised) version based on the ancient U-Boot 2009.11
> Other commands are also effected, for example 'ide dev 0' would abort a 'run' command if the device did not exist.
> 
> Is this my problem (corrupted source\compilation) or a change in policy ?

I took a quick glance over command/main.c. I think this is the "original
behaviour" (tree 9c506e 23 Aug 2011):

1374                 /* OK - call function to do the command */
1375                 if ((cmdtp->cmd) (cmdtp, flag, argc, argv) != 0) {
1376                         rc = -1;
1377                 }

run_command returns -1 on failure and

1407                 if (run_command (arg, flag) == -1)
1408                         return 1;

do_run exits the loop based on that.

Now (tree 762494 6 Mar 2012) builtin_run_command:

1341                 rc = cmd_process(flag, argc, argv, &repeatable);

returns the exit code of the command but:

1366         if (builtin_run_command(cmd, flag) == -1)
1367                 return 1;

run_command now depends on it to return -1 on failure. If I followed the
code right, this is clear change in the behaviour. I hope it is not
intentional and can be fixed because I have (too?) scripts that depends
on old behaviour. For example:

run get_update_from_usb erase_flash_and_write_it

Sequence must not continue to touching flash if load from USB fails.

--

Timo