Re: remote.exp: fail properly if wrapper status message isn't found

Re: remote.exp: fail properly if wrapper status message isn't found

[Andrew Cagney]

Should this be committed to the src/dejagnu source base (regardless of dejagnu)?


Andrew

> Hi.
> 
> I had a bug in the fini code in the newlib port of
> mmix-knuth-mmixware (film at 11) which exposed this dejagnu
> fault: a target using a wrapper gets a PASS if the "*** EXIT
> code ..." pattern isn't output.  IIRC, at the time I tested
> this on other simulator targets, that fault exposed two other
> gcc testsuite failures on some other target.  I can't remember
> which; it might have been d30v or mn10300.
> 
> Please consider applying this.  Please make sure to CC me on
> follow-ups as I'm not subscribed to these lists.
> 
> 2001-11-10  Hans-Peter Nilsson  <hp@bitrange.com>
> 
> * lib/remote.exp (check_for_board_status): Return 2 if the board
> 	uses a status wrapper and there was no match for the expected
> 	pattern.
> 
> Index: remote.exp
> ===================================================================
> RCS file: /cvsroot/dejagnu/dejagnu/lib/remote.exp,v
> retrieving revision 1.4
> diff -p -c -r1.4 remote.exp
> *** remote.exp	2001/10/02 15:48:51	1.4
> --- remote.exp	2001/11/10 15:59:40
> *************** proc standard_wait { dest timeout } {
> *** 1050,1055 ****
> --- 1050,1065 ----
>   proc check_for_board_status  { variable } {
>       upvar $variable output;
> 
> +     # If all programs of this board have a wrapper that always outputs a
> +     # status message, then the absence of it means that the program
> +     # crashed, regardless of status found elsewhere (e.g. simulator exit
> +     # code).
> +     if { [target_info needs_status_wrapper] != "" } then {
> + 	set nomatch_return 2
> +     } else {
> + 	set nomatch_return -1
> +     }
> +
>       if [regexp "(^|\[\r\n\])\\*\\*\\* EXIT code" $output] {
>   	regsub "^.*\\*\\*\\* EXIT code " $output "" result;
>   	regsub "\[\r\n\].*$" $result "" result;
> *************** proc check_for_board_status  { variable
> *** 1059,1070 ****
>   	verbose "got board status $result" 3
>   	verbose "output is $output" 3
>   	if { $result == "" } {
> ! 	    return -1;
>   	} else {
>   	    return [expr $result];
>   	}
>       } else {
> ! 	return -1;
>       }
>   }
> 
> --- 1069,1080 ----
>   	verbose "got board status $result" 3
>   	verbose "output is $output" 3
>   	if { $result == "" } {
> ! 	    return $nomatch_return
>   	} else {
>   	    return [expr $result];
>   	}
>       } else {
> ! 	return $nomatch_return
>       }
>   }
> 
> 

Re: remote.exp: fail properly if wrapper status message isn't found

[Fernando Nasser]

Andrew Cagney wrote:
> 
> Should this be committed to the src/dejagnu source base (regardless of dejagnu)?
> 
> Andrew
> 

We'd normally do that _and_ send the patch to the dejagnu external
maintainer.

Fernando

> > Hi.
> >
> > I had a bug in the fini code in the newlib port of
> > mmix-knuth-mmixware (film at 11) which exposed this dejagnu
> > fault: a target using a wrapper gets a PASS if the "*** EXIT
> > code ..." pattern isn't output.  IIRC, at the time I tested
> > this on other simulator targets, that fault exposed two other
> > gcc testsuite failures on some other target.  I can't remember
> > which; it might have been d30v or mn10300.
> >
> > Please consider applying this.  Please make sure to CC me on
> > follow-ups as I'm not subscribed to these lists.
> >
> > 2001-11-10  Hans-Peter Nilsson  <hp@bitrange.com>
> >
> > * lib/remote.exp (check_for_board_status): Return 2 if the board
> >       uses a status wrapper and there was no match for the expected
> >       pattern.
> >
> > Index: remote.exp
> > ===================================================================
> > RCS file: /cvsroot/dejagnu/dejagnu/lib/remote.exp,v
> > retrieving revision 1.4
> > diff -p -c -r1.4 remote.exp
> > *** remote.exp        2001/10/02 15:48:51     1.4
> > --- remote.exp        2001/11/10 15:59:40
> > *************** proc standard_wait { dest timeout } {
> > *** 1050,1055 ****
> > --- 1050,1065 ----
> >   proc check_for_board_status  { variable } {
> >       upvar $variable output;
> >
> > +     # If all programs of this board have a wrapper that always outputs a
> > +     # status message, then the absence of it means that the program
> > +     # crashed, regardless of status found elsewhere (e.g. simulator exit
> > +     # code).
> > +     if { [target_info needs_status_wrapper] != "" } then {
> > +     set nomatch_return 2
> > +     } else {
> > +     set nomatch_return -1
> > +     }
> > +
> >       if [regexp "(^|\[\r\n\])\\*\\*\\* EXIT code" $output] {
> >       regsub "^.*\\*\\*\\* EXIT code " $output "" result;
> >       regsub "\[\r\n\].*$" $result "" result;
> > *************** proc check_for_board_status  { variable
> > *** 1059,1070 ****
> >       verbose "got board status $result" 3
> >       verbose "output is $output" 3
> >       if { $result == "" } {
> > !         return -1;
> >       } else {
> >           return [expr $result];
> >       }
> >       } else {
> > !     return -1;
> >       }
> >   }
> >
> > --- 1069,1080 ----
> >       verbose "got board status $result" 3
> >       verbose "output is $output" 3
> >       if { $result == "" } {
> > !         return $nomatch_return
> >       } else {
> >           return [expr $result];
> >       }
> >       } else {
> > !     return $nomatch_return
> >       }
> >   }
> >
> >

-- 
Fernando Nasser
Red Hat Canada Ltd.                     E-Mail:  fnasser@redhat.com
2323 Yonge Street, Suite #300
Toronto, Ontario   M4P 2C9

Re: remote.exp: fail properly if wrapper status message isn't found

[Andrew Cagney]

> Andrew Cagney wrote:
> 
>> 
>> Should this be committed to the src/dejagnu source base (regardless of dejagnu)?
>> 
>> Andrew
>> 
> 
> 
> We'd normally do that _and_ send the patch to the dejagnu external
> maintainer.


Just FYI, Hans-Peter's original post went to both lists.

Andrew