On 3/24/26 16:20, Joseph Myers wrote:
There is a bogus comment "We use && here, as otherwise the echo always
works, which makes it look like execution succeeded when in reality it
failed.". But the whole point of the XYZ${?}ZYX logic, taken from
rsh.exp, is that code a few lines below uses a regular expression to
extract the actual exit status from the test output, which the use of
'&&' breaks when the test program does fail. Furthermore, if the
intent was to yield nonzero exit status from local_exec when the test
program fails, the subsequent rm would prevent that from working by
yielding the exit status from rm instead.
---
lib/ssh.exp | 4 +---
1 file changed, 1 insertion(+), 3 deletions(-)
diff --git a/lib/ssh.exp b/lib/ssh.exp
index bb827f4..b64bf73 100644
--- a/lib/ssh.exp
+++ b/lib/ssh.exp
@@ -168,9 +168,7 @@ proc ssh_exec { boardname program pargs inp outp } {
set inp "/dev/null"
}
- # We use && here, as otherwise the echo always works, which makes it look
- # like execution succeeded when in reality it failed.
- set ret [local_exec "$SSH $ssh_useropts $ssh_user$hostname sh -c '$program $pargs 2>&1
&& echo XYZ\\\${?}ZYX \\; rm -f $program'" $inp $outp $timeout]
+ set ret [local_exec "$SSH $ssh_useropts $ssh_user$hostname sh -c '$program $pargs
2>&1 ; echo XYZ\\\${?}ZYX \\; rm -f $program'" $inp $outp $timeout]
set status [lindex $ret 0]
set output [lindex $ret 1]
This corrects a logic error; applied.
The echo(1) is used to report the exit status of the program, but using
&& means that the report only appears if the program completes
successfully, which is ridiculous because an error is exactly when you
most want the exit code.
-- Jacob
