[PATCH] Add -fchecking

[PATCH] Add -fchecking

[Richard Biener]

This adds -fchecking as a way to enable internal consistency checks
even in release builds (or disable checking with -fno-checking - up to
a certain extent - with checking enabled).

Bootstrap & regtest running on x86_64-unknown-linux-gnu.

Richard.

2015-10-27  Richard Biener  <rguenther@suse.de>

	* common.opt (fchecking): New flag controlling flag_checking.
	* passes.c (execute_function_todo): Guard checking code with
	flag_checking instead of ENABLE_CHECKING.
	(execute_todo): Likewise.
	(execute_one_pass): Likewise.
	(verify_curr_properties): Always compile.
	* timevar.c: Include options.h.
	(timer::print): Adjust output.
	* doc/invoke.texi (fchecking): Document.

Index: gcc/common.opt
===================================================================
*** gcc/common.opt	(revision 229404)
--- gcc/common.opt	(working copy)
*************** int optimize_fast
*** 46,56 ****
  Variable
  bool in_lto_p = false
  
- ; Enable additional checks of internal state consistency, which may slow
- ; the compiler down.
- Variable
- bool flag_checking = CHECKING_P
- 
  ; 0 means straightforward implementation of complex divide acceptable.
  ; 1 means wide ranges of inputs must work for complex divide.
  ; 2 means C99-like requirements for complex multiply and divide.
--- 46,51 ----
*************** fcheck-new
*** 1002,1007 ****
--- 997,1006 ----
  Common Var(flag_check_new)
  Check the return value of new in C++.
  
+ fchecking
+ Common Var(flag_checking) Init(CHECKING_P)
+ Perform internal consistency checkings.
+ 
  fcombine-stack-adjustments
  Common Report Var(flag_combine_stack_adjustments) Optimization
  Looks for opportunities to reduce stack adjustments and stack references.
Index: gcc/passes.c
===================================================================
*** gcc/passes.c	(revision 229404)
--- gcc/passes.c	(working copy)
*************** execute_function_todo (function *fn, voi
*** 1952,1960 ****
  
    gcc_assert (dom_info_state (fn, CDI_POST_DOMINATORS) == DOM_NONE);
    /* If we've seen errors do not bother running any verifiers.  */
!   if (!seen_error ())
      {
- #if defined ENABLE_CHECKING
        dom_state pre_verify_state = dom_info_state (fn, CDI_DOMINATORS);
        dom_state pre_verify_pstate = dom_info_state (fn, CDI_POST_DOMINATORS);
  
--- 1952,1960 ----
  
    gcc_assert (dom_info_state (fn, CDI_POST_DOMINATORS) == DOM_NONE);
    /* If we've seen errors do not bother running any verifiers.  */
!   if (!seen_error ()
!       && flag_checking)
      {
        dom_state pre_verify_state = dom_info_state (fn, CDI_DOMINATORS);
        dom_state pre_verify_pstate = dom_info_state (fn, CDI_POST_DOMINATORS);
  
*************** execute_function_todo (function *fn, voi
*** 1988,1994 ****
        /* Make sure verifiers don't change dominator state.  */
        gcc_assert (dom_info_state (fn, CDI_DOMINATORS) == pre_verify_state);
        gcc_assert (dom_info_state (fn, CDI_POST_DOMINATORS) == pre_verify_pstate);
- #endif
      }
  
    fn->last_verified = flags & TODO_verify_all;
--- 1991,1996 ----
*************** execute_function_todo (function *fn, voi
*** 2008,2018 ****
  static void
  execute_todo (unsigned int flags)
  {
! #if defined ENABLE_CHECKING
!   if (cfun
        && need_ssa_update_p (cfun))
      gcc_assert (flags & TODO_update_ssa_any);
- #endif
  
    timevar_push (TV_TODO);
  
--- 2010,2019 ----
  static void
  execute_todo (unsigned int flags)
  {
!   if (flag_checking
!       && cfun
        && need_ssa_update_p (cfun))
      gcc_assert (flags & TODO_update_ssa_any);
  
    timevar_push (TV_TODO);
  
*************** clear_last_verified (function *fn, void
*** 2076,2089 ****
  /* Helper function. Verify that the properties has been turn into the
     properties expected by the pass.  */
  
- #ifdef ENABLE_CHECKING
  static void
  verify_curr_properties (function *fn, void *data)
  {
    unsigned int props = (size_t)data;
    gcc_assert ((fn->curr_properties & props) == props);
  }
- #endif
  
  /* Initialize pass dump file.  */
  /* This is non-static so that the plugins can use it.  */
--- 2077,2088 ----
*************** execute_one_pass (opt_pass *pass)
*** 2331,2344 ****
    /* Run pre-pass verification.  */
    execute_todo (pass->todo_flags_start);
  
! #ifdef ENABLE_CHECKING
!   do_per_function (verify_curr_properties,
! 		   (void *)(size_t)pass->properties_required);
! #endif
  
    /* If a timevar is present, start it.  */
    if (pass->tv_id != TV_NONE)
!     timevar_push (pass->tv_id);
  
    /* Do it!  */
    todo_after = pass->execute (cfun);
--- 2330,2342 ----
    /* Run pre-pass verification.  */
    execute_todo (pass->todo_flags_start);
  
!   if (flag_checking)
!     do_per_function (verify_curr_properties,
! 		     (void *)(size_t)pass->properties_required);
  
    /* If a timevar is present, start it.  */
    if (pass->tv_id != TV_NONE)
!     timevar_push (pass->tv_id);
  
    /* Do it!  */
    todo_after = pass->execute (cfun);

Index: gcc/doc/invoke.texi
===================================================================
*** gcc/doc/invoke.texi	(revision 229404)
--- gcc/doc/invoke.texi	(working copy)
*************** Objective-C and Objective-C++ Dialects}.
*** 310,316 ****
  -fsanitize=@var{style} -fsanitize-recover -fsanitize-recover=@var{style} @gol
  -fasan-shadow-offset=@var{number} -fsanitize-sections=@var{s1},@var{s2},... @gol
  -fsanitize-undefined-trap-on-error @gol
! -fcheck-pointer-bounds -fchkp-check-incomplete-type @gol
  -fchkp-first-field-has-own-bounds -fchkp-narrow-bounds @gol
  -fchkp-narrow-to-innermost-array -fchkp-optimize @gol
  -fchkp-use-fast-string-functions -fchkp-use-nochk-string-functions @gol
--- 310,316 ----
  -fsanitize=@var{style} -fsanitize-recover -fsanitize-recover=@var{style} @gol
  -fasan-shadow-offset=@var{number} -fsanitize-sections=@var{s1},@var{s2},... @gol
  -fsanitize-undefined-trap-on-error @gol
! -fcheck-pointer-bounds -fchecking -fchkp-check-incomplete-type @gol
  -fchkp-first-field-has-own-bounds -fchkp-narrow-bounds @gol
  -fchkp-narrow-to-innermost-array -fchkp-optimize @gol
  -fchkp-use-fast-string-functions -fchkp-use-nochk-string-functions @gol
*************** structure field attribute (@pxref{Type A
*** 6145,6150 ****
--- 6145,6156 ----
  functions for controlling the Pointer Bounds Checker.  @xref{Pointer
  Bounds Checker builtins}, for more information.
  
+ @item -fchecking
+ @opindex fchecking
+ @opindex fno-checking
+ Enable internal consistency checking.  The default depends on
+ the compiler configuration.
+ 
  @item -fchkp-check-incomplete-type
  @opindex fchkp-check-incomplete-type
  @opindex fno-chkp-check-incomplete-type
Index: gcc/timevar.c
===================================================================
*** gcc/timevar.c	(revision 229404)
--- gcc/timevar.c	(working copy)
*************** along with GCC; see the file COPYING3.
*** 22,27 ****
--- 22,28 ----
  #include "system.h"
  #include "coretypes.h"
  #include "timevar.h"
+ #include "options.h"
  
  #ifndef HAVE_CLOCK_T
  typedef int clock_t;
*************** timer::print (FILE *fp)
*** 727,736 ****
  #endif
    fprintf (fp, "%8u kB\n", (unsigned) (total->ggc_mem >> 10));
  
! #ifdef ENABLE_CHECKING
!   fprintf (fp, "Extra diagnostic checks enabled; compiler may run slowly.\n");
!   fprintf (fp, "Configure with --enable-checking=release to disable checks.\n");
! #endif
  #ifndef ENABLE_ASSERT_CHECKING
    fprintf (fp, "Internal checks disabled; compiler is not suited for release.\n");
    fprintf (fp, "Configure with --enable-checking=release to enable checks.\n");
--- 728,737 ----
  #endif
    fprintf (fp, "%8u kB\n", (unsigned) (total->ggc_mem >> 10));
  
!   if (CHECKING_P || flag_checking)
!     fprintf (fp, "Extra diagnostic checks enabled; compiler may run slowly.\n");
!   if (CHECKING_P)
!     fprintf (fp, "Configure with --enable-checking=release to disable checks.\n");
  #ifndef ENABLE_ASSERT_CHECKING
    fprintf (fp, "Internal checks disabled; compiler is not suited for release.\n");
    fprintf (fp, "Configure with --enable-checking=release to enable checks.\n");

Re: [PATCH] Add -fchecking

[Mikhail Maltsev]

On 10/27/2015 04:17 PM, Richard Biener wrote:
> 
> This adds -fchecking as a way to enable internal consistency checks
> even in release builds (or disable checking with -fno-checking - up to
> a certain extent - with checking enabled).

I remember that Jakub proposed to use __builtin_expect with
flag_checking. I wonder, if it is possible to implement without hacking
AWK scripts just for this particular case? For example, to define
flag_checking to something like

#define flag_checking __builtin_expect (flag_checking_val, CHECKING_P)

(provided that flag_checking_val is the actual value got from
command-line options).

-- 
Regards,
    Mikhail Maltsev

Re: [PATCH] Add -fchecking

[Jeff Law]

On 10/27/2015 09:32 AM, Mikhail Maltsev wrote:
> On 10/27/2015 04:17 PM, Richard Biener wrote:
>>
>> This adds -fchecking as a way to enable internal consistency checks
>> even in release builds (or disable checking with -fno-checking - up to
>> a certain extent - with checking enabled).
>
> I remember that Jakub proposed to use __builtin_expect with
> flag_checking. I wonder, if it is possible to implement without hacking
> AWK scripts just for this particular case? For example, to define
> flag_checking to something like
>
> #define flag_checking __builtin_expect (flag_checking_val, CHECKING_P)
>
> (provided that flag_checking_val is the actual value got from
> command-line options).
I think this ought to be a follow-up item.  And yes, we're going to need 
some level of indirection so that we're not writing __builtin_expect all 
over the place.

Jeff

Re: [PATCH] Add -fchecking

[Richard Biener]

On Tue, 27 Oct 2015, Jeff Law wrote:

> On 10/27/2015 09:32 AM, Mikhail Maltsev wrote:
> > On 10/27/2015 04:17 PM, Richard Biener wrote:
> > > 
> > > This adds -fchecking as a way to enable internal consistency checks
> > > even in release builds (or disable checking with -fno-checking - up to
> > > a certain extent - with checking enabled).
> > 
> > I remember that Jakub proposed to use __builtin_expect with
> > flag_checking. I wonder, if it is possible to implement without hacking
> > AWK scripts just for this particular case? For example, to define
> > flag_checking to something like
> > 
> > #define flag_checking __builtin_expect (flag_checking_val, CHECKING_P)
> > 
> > (provided that flag_checking_val is the actual value got from
> > command-line options).
> I think this ought to be a follow-up item.  And yes, we're going to need some
> level of indirection so that we're not writing __builtin_expect all over the
> place.

Yeah, we should do that as followup.  We should also make sure to
only make the branch unlikely and not end up with optimizing
all checking code for size (making it even slower than it is now).

Richard.

> Jeff
> 
> 

-- 
Richard Biener <rguenther@suse.de>
SUSE LINUX GmbH, GF: Felix Imendoerffer, Jane Smithard, Graham Norton, HRB 21284 (AG Nuernberg)

Re: [PATCH] Add -fchecking

[Gerald Pfeifer]

On Tue, 27 Oct 2015, Richard Biener wrote:
> This adds -fchecking as a way to enable internal consistency checks
> even in release builds (or disable checking with -fno-checking - up to
> a certain extent - with checking enabled).

How (much) do we want to advertize this?

Gerald

Re: [PATCH] Add -fchecking

[Jeff Law]

On 11/07/2015 01:47 PM, Gerald Pfeifer wrote:
> On Tue, 27 Oct 2015, Richard Biener wrote:
>> This adds -fchecking as a way to enable internal consistency checks
>> even in release builds (or disable checking with -fno-checking - up to
>> a certain extent - with checking enabled).
>
> How (much) do we want to advertize this?
I don't think much -- it's really a developer-centric option.

Jeff

Re: [PATCH] Add -fchecking

[Richard Biener]

On November 8, 2015 1:40:57 AM GMT+01:00, Jeff Law <law@redhat.com> wrote:
>On 11/07/2015 01:47 PM, Gerald Pfeifer wrote:
>> On Tue, 27 Oct 2015, Richard Biener wrote:
>>> This adds -fchecking as a way to enable internal consistency checks
>>> even in release builds (or disable checking with -fno-checking - up
>to
>>> a certain extent - with checking enabled).
>>
>> How (much) do we want to advertize this?
>I don't think much -- it's really a developer-centric option.

Might be worth mentioning on bugs.html as sometimes miscompiles can be detected by -fchecking.

Richard.

>Jeff