[m-dev.] for review: typeinfo-liveness bug fix

[Mark Anthony BROWN]

Zoltan Somogyi writes:
> For review by anyone.
> 
> compiler/liveness.m:
> 	Fix a bug: the compiler did not make sure that the sets of variables
> 	in resume_points satisfy the requirements of typeinfo liveness when
> 	required (i.e. it included variables in the resume point set without
> 	also including the typeinfos describing their types). This lead to a
> 	compiler abort when compiling with options (e.g. --trace deep) that
> 	imply typeinfo liveness.
> 
> tests/debugger/resume_typeinfos.{m,inp,exp,exp2}
> 	A test case to exercise the fix.
> 
> tests/debugger/Mmakefile:
> 	Enable the new test case, and put the test list back into alphabetic
> 	order.
> 
> Zoltan.
> 

Apart from the comments below, this change looks fine.

Cheers,
Mark.


> @@ -777,6 +783,14 @@
>  	% we stop generating code after the first cannot_fail disjunct.
>  	% Second, an empty pruned disjunction is legal, while an empty
>  	% nondet disjunction isn't.
> +	%
> +	% For both kinds of disjunctions, the resume points to be attached to
> +	% the non-last disjuncts must be completed with the required typeinfos
> +	% if --typeinfo-liveness is set. ResumeVars0 should already be so
> +	% completed, so need only completed the sets added here.

That sentence doesn't quite make sense.


> Index: tests/debugger/Mmakefile
> ===================================================================
> RCS file: /home/mercury1/repository/tests/debugger/Mmakefile,v
> retrieving revision 1.41
> diff -u -b -r1.41 Mmakefile
> --- tests/debugger/Mmakefile	2000/06/19 07:59:18	1.41
> +++ tests/debugger/Mmakefile	2000/07/17 22:52:48
> @@ -30,9 +30,10 @@
>  	loopcheck			\
>  	mdb_command_test		\
>  	multi_parameter			\
> +	polymorphic_output		\
>  	queens				\
> -	shallow				\
> -	polymorphic_output
> +	resume_typeinfos		\
> +	shallow

Doesn't this test require a special rule in the Mmakefile, like the
other tests in this directory?

>  
>  # The following tests are disabled, since currently they get some spurious
>  # failures if readline support is enabled:
> Index: tests/debugger/resume_typeinfos.exp
> ===================================================================
> RCS file: resume_typeinfos.exp
> diff -N resume_typeinfos.exp
> --- /dev/null	Thu Mar  4 04:20:11 1999
> +++ resume_typeinfos.exp	Mon Jul 17 18:40:42 2000
> @@ -0,0 +1,14 @@
> +       1:      1  1 CALL pred resume_typeinfos:main/2-0 (det) resume_typeinfos.m:25
> +mdb> echo on
> +Command echo enabled.
> +mdb> context none
> +Contexts will not be printed.
> +mdb> goto 5
> +       5:      4  3 CALL pred list:length/2-0 (det)
> +mdb> finish
> +      15:      4  3 EXIT pred list:length/2-0 (det)
> +mdb> print *
> +       HeadVar__1             	[1, 2]
> +       HeadVar__2             	2
> +mdb> continue
> +no solution.
> Index: tests/debugger/resume_typeinfos.exp2
> ===================================================================
> RCS file: resume_typeinfos.exp2
> diff -N resume_typeinfos.exp2
> --- /dev/null	Thu Mar  4 04:20:11 1999
> +++ resume_typeinfos.exp2	Mon Jul 17 18:40:42 2000
> @@ -0,0 +1,14 @@
> +       1:      1  1 CALL pred resume_typeinfos:main/2-0 (det) resume_typeinfos.m:25
> +mdb> echo on
> +Command echo enabled.
> +mdb> context none
> +Contexts will not be printed.
> +mdb> goto 5
> +       5:      4  3 CALL pred list:length/2-0 (det)
> +mdb> finish
> +      15:      4  3 EXIT pred list:length/2-0 (det)
> +mdb> print *
> +       HeadVar__1             	[1, 2]
> +       HeadVar__2             	2
> +mdb> continue
> +no solution.

Is there any difference in the above two files?  If not, why is the
second one required?

--------------------------------------------------------------------------
mercury-developers mailing list
Post messages to:       mercury-developers at cs.mu.oz.au
Administrative Queries: owner-mercury-developers at cs.mu.oz.au
Subscriptions:          mercury-developers-request at cs.mu.oz.au
--------------------------------------------------------------------------