[Bug bootstrap/103899] New: make profiledbootstrap fails due to uninitialized warning in expr.c

[Bug bootstrap/103899] New: make profiledbootstrap fails due to uninitialized warning in expr.c

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103899

            Bug ID: 103899
           Summary: make profiledbootstrap fails due to uninitialized
                    warning in expr.c
           Product: gcc
           Version: 12.0
            Status: UNCONFIRMED
          Keywords: build, diagnostic
          Severity: normal
          Priority: P3
         Component: bootstrap
          Assignee: unassigned at gcc dot gnu.org
          Reporter: pinskia at gcc dot gnu.org
  Target Milestone: ---

I am filing this and will attach the preprocessed source in a few minutes:
/home/apinski/src/upstream-gcc/gcc/objdir1/./prev-gcc/xg++
-B/home/apinski/src/upstream-gcc/gcc/objdir1/./prev-gcc/
-B/usr/local/x86_64-pc-linux-gnu/bin/ -nostdinc++
-B/home/apinski/src/upstream-gcc/gcc/objdir1/prev-x86_64-pc-linux-gnu/libstdc++-v3/src/.libs
-B/home/apinski/src/upstream-gcc/gcc/objdir1/prev-x86_64-pc-linux-gnu/libstdc++-v3/libsupc++/.libs

-I/home/apinski/src/upstream-gcc/gcc/objdir1/prev-x86_64-pc-linux-gnu/libstdc++-v3/include/x86_64-pc-linux-gnu

-I/home/apinski/src/upstream-gcc/gcc/objdir1/prev-x86_64-pc-linux-gnu/libstdc++-v3/include
 -I/home/apinski/src/upstream-gcc/gcc/libstdc++-v3/libsupc++
-L/home/apinski/src/upstream-gcc/gcc/objdir1/prev-x86_64-pc-linux-gnu/libstdc++-v3/src/.libs
-L/home/apinski/src/upstream-gcc/gcc/objdir1/prev-x86_64-pc-linux-gnu/libstdc++-v3/libsupc++/.libs
 -fno-PIE -c   -g -O2 -fno-checking -gtoggle -fprofile-generate -DIN_GCC    
-fno-exceptions -fno-rtti -fasynchronous-unwind-tables -W -Wall -Wno-narrowing
-Wwrite-strings -Wcast-qual -Wno-error=format-diag -Wmissing-format-attribute
-Woverloaded-virtual -pedantic -Wno-long-long -Wno-variadic-macros
-Wno-overlength-strings -Werror -fno-common  -DHAVE_CONFIG_H -I. -I.
-I../../gcc -I../../gcc/. -I../../gcc/../include -I../../gcc/../libcpp/include
-I../../gcc/../libcody -I/home/apinski/src/upstream-gcc/gcc/objdir1/./gmp
-I/home/apinski/src/upstream-gcc/gcc/gmp
-I/home/apinski/src/upstream-gcc/gcc/objdir1/./mpfr/src
-I/home/apinski/src/upstream-gcc/gcc/mpfr/src
-I/home/apinski/src/upstream-gcc/gcc/mpc/src  -I../../gcc/../libdecnumber
-I../../gcc/../libdecnumber/bid -I../libdecnumber -I../../gcc/../libbacktrace
-I/home/apinski/src/upstream-gcc/gcc/objdir1/./isl/include
-I/home/apinski/src/upstream-gcc/gcc/isl/include  -o expr.o -MT expr.o -MMD -MP
-MF ./.deps/expr.TPo ../../gcc/expr.c
In file included from ../../gcc/expr.c:26:
../../gcc/tree.h: In function ‘rtx_def* expand_expr_real_1(tree, rtx,
machine_mode, expand_modifier, rtx_def**, bool)’:
../../gcc/tree.h:244:56: error: ‘context’ may be used uninitialized in this
function [-Werror=maybe-uninitialized]
  244 | #define TREE_CODE(NODE) ((enum tree_code) (NODE)->base.code)
      |                                                        ^~~~
../../gcc/expr.c:10343:8: note: ‘context’ was declared here
10343 |   tree context;
      |        ^~~~~~~
cc1plus: all warnings being treated as errors
Makefile:1143: recipe for target 'expr.o' failed
make[3]: *** [expr.o] Error 1

I just did:
mkdir objdir1
cd objdir1
../configure
make profiledbootstrap

I tried to get a reduced testcase but I think there is a missing jump threading
so it just happen to work for the reduced testcase I tried.

[Bug bootstrap/103899] make profiledbootstrap fails due to uninitialized warning in expr.c

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103899

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Created attachment 52113
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=52113&action=edit
unreduced testcase

../prev-gcc/cc1plus expr-warning.ii -fpreprocessed -fno-PIE   -g -O2
-fno-checking -gtoggle -fprofile-generate   -fno-exceptions -fno-rtti
-fasynchronous-unwind-tables -W -Wall -Wno-narrowing -Wwrite-strings
-Wcast-qual -Wno-error=format-diag -Wmissing-format-attribute
-Woverloaded-virtual -pedantic -Wno-long-long -Wno-variadic-macros
-Wno-overlength-strings -Werror -fno-common -Wno-implicit-fallthrough -quiet

[Bug bootstrap/103899] make profiledbootstrap fails due to uninitialized warning in expr.c

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103899

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
I am going to try to reduce the issue as I know without -fprofile-generate, GCC
does not warn .

[Bug bootstrap/103899] make profiledbootstrap fails due to uninitialized warning in expr.c

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103899

--- Comment #3 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
I suspect this is a regression and one caused by some of the jump threading
patches. Let me finish up reducing the testcase and I will test the reduced one
on different versions of GCC.

[Bug tree-optimization/103899] [12 Regression] make profiledbootstrap fails due to uninitialized warning in expr.c

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103899

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |missed-optimization,
                   |                            |needs-bisection
               Host|x86_64-linux-gnu            |
          Component|bootstrap                   |tree-optimization
              Build|x86_64-linux-gnu            |
            Summary|make profiledbootstrap      |[12 Regression] make
                   |fails due to uninitialized  |profiledbootstrap fails due
                   |warning in expr.c           |to uninitialized warning in
                   |                            |expr.c
   Target Milestone|---                         |12.0

--- Comment #4 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Reduced testcase:
int h;
int l ();
int *k ();
void nn ();
void expand_expr_real_1 (int *exp)
{
        int m;
        int *context;
        m = 2;
        if (h || !exp)
                m = l ();
        if (exp)
                context = k ();
        if (exp && context && context[0] == 0)
        if (m == 0)
           nn ();
}
----- CUT ----
Compile with "-O2 -W -Wall -Werror", there is no warning but once you add
-fprofile-generate there is an uninitialized warning for context.
There is a missing jump threading on the trunk which is causing the warning to
show up.

[Bug tree-optimization/103899] [12 Regression] make profiledbootstrap fails due to uninitialized warning in expr.c

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103899

Martin Liška <marxin at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever confirmed|0                           |1
   Last reconfirmed|                            |2022-01-05
                 CC|                            |aldyh at gcc dot gnu.org

--- Comment #5 from Martin Liška <marxin at gcc dot gnu.org> ---
(In reply to Andrew Pinski from comment #4)
> Reduced testcase:
> int h;
> int l ();
> int *k ();
> void nn ();
> void expand_expr_real_1 (int *exp)
> {
> 	int m;
> 	int *context;
> 	m = 2;
> 	if (h || !exp)
> 		m = l ();
> 	if (exp)
> 		context = k ();
> 	if (exp && context && context[0] == 0)
> 	if (m == 0)
> 	   nn ();
> }
> ----- CUT ----
> Compile with "-O2 -W -Wall -Werror", there is no warning but once you add
> -fprofile-generate there is an uninitialized warning for context.
> There is a missing jump threading on the trunk which is causing the warning
> to show up.

Started with r12-4790-g4b3a325f07acebf4, may I remove needs-bisection Keyword?

[Bug tree-optimization/103899] [12 Regression] make profiledbootstrap fails due to uninitialized warning in expr.c

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103899

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jakub at gcc dot gnu.org

--- Comment #6 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
While we should fix uninit warning to deal with this if possible, I think we
should also try to work around this in expr.c.

If following works, then I think it would be even a cleanup (as it moves the
context variable declaration to the sole spot where it is used):

--- gcc/expr.c.jj       2022-01-03 10:40:41.203164211 +0100
+++ gcc/expr.c  2022-01-05 14:47:47.684660031 +0100
@@ -10340,7 +10340,6 @@ expand_expr_real_1 (tree exp, rtx target
   enum tree_code code = TREE_CODE (exp);
   rtx subtarget, original_target;
   int ignore;
-  tree context;
   bool reduce_bit_field;
   location_t loc = EXPR_LOCATION (exp);
   struct separate_ops ops;
@@ -10579,14 +10578,16 @@ expand_expr_real_1 (tree exp, rtx target
       /* Variables inherited from containing functions should have
         been lowered by this point.  */
       if (exp)
-       context = decl_function_context (exp);
-      gcc_assert (!exp
-                 || SCOPE_FILE_SCOPE_P (context)
-                 || context == current_function_decl
-                 || TREE_STATIC (exp)
-                 || DECL_EXTERNAL (exp)
-                 /* ??? C++ creates functions that are not TREE_STATIC.  */
-                 || TREE_CODE (exp) == FUNCTION_DECL);
+       {
+         tree context = decl_function_context (exp);
+         gcc_assert (SCOPE_FILE_SCOPE_P (context)
+                     || context == current_function_decl
+                     || TREE_STATIC (exp)
+                     || DECL_EXTERNAL (exp)
+                     /* ??? C++ creates functions that are not
+                        TREE_STATIC.  */
+                     || TREE_CODE (exp) == FUNCTION_DECL);
+       }

       /* This is the case of an array whose size is to be determined
         from its initializer, while the initializer is still being parsed.

[Bug tree-optimization/103899] [12 Regression] make profiledbootstrap fails due to uninitialized warning in expr.c

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103899

--- Comment #7 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Confirmed it does help.

[Bug tree-optimization/103899] [12 Regression] make profiledbootstrap fails due to uninitialized warning in expr.c

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103899

--- Comment #8 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Jakub Jelinek from comment #6)
> While we should fix uninit warning to deal with this if possible, I think we
> should also try to work around this in expr.c.
> 
> If following works, then I think it would be even a cleanup (as it moves the
> context variable declaration to the sole spot where it is used):

Yes I do think we should improve the code in expr.c to better express this and
have this cleanup.

[Bug tree-optimization/103899] [12 Regression] make profiledbootstrap fails due to uninitialized warning in expr.c

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103899

--- Comment #9 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Jakub Jelinek <jakub@gcc.gnu.org>:

https://gcc.gnu.org/g:db33b1059bcee3eeb36d174d59abf19f2bef5d66

commit r12-6280-gdb33b1059bcee3eeb36d174d59abf19f2bef5d66
Author: Jakub Jelinek <jakub@redhat.com>
Date:   Thu Jan 6 15:32:57 2022 +0100

    expr: Workaround profiledbootstrap uninit false positive [PR103899]

    The threader changes resulted in a false positive warning during
    profiledbootstrap:
    In file included from ../../gcc/expr.c:26:
    ../../gcc/tree.h: In function ârtx_def* expand_expr_real_1(tree, rtx,
machine_mode, expand_modifier, rtx_def**, bool)â:
    ../../gcc/tree.h:244:56: error: âcontextâ may be used uninitialized in
this function [-Werror=maybe-uninitialized]
      244 | #define TREE_CODE(NODE) ((enum tree_code) (NODE)->base.code)
          |                                                        ^~~~
    ../../gcc/expr.c:10343:8: note: âcontextâ was declared here
    10343 |   tree context;
          |        ^~~~~~~
    While it will be nice to improve the uninit pass to handle it if possible
    (I do not want to close the PR until that is done), doing profiledbootstrap
    is a common thing to do, so a workaround is handy, especially as in this
    case when the workaround seems to be the right thing to do, as it moves
    a variable declaration to the only place where it is set and used and
avoids
    the weird and for uninit asking
      tree context;
    ...
      if (exp)
        context = ...;
      gcc_assert (!exp
                  || use (context)
                  || use_some_more (context));

    2022-01-06  Jakub Jelinek  <jakub@redhat.com>

            PR tree-optimization/103899
            * expr.c (expand_expr_real_1): Add a workaround for bogus uninit
            warning by moving context variable to the only spot where it is
used
            and moving gcc_assert into if body.

[Bug tree-optimization/103899] [12 Regression] make profiledbootstrap fails due to uninitialized warning in expr.c

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103899

--- Comment #10 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Workaround applied, tree-ssa-uninit.c should be improved if possible though.

[Bug tree-optimization/103899] [12 Regression] make profiledbootstrap fails due to uninitialized warning in expr.c

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103899

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |FIXED
             Status|NEW                         |RESOLVED

--- Comment #11 from Richard Biener <rguenth at gcc dot gnu.org> ---
The build regression is fixed.  If there's interest please file the comment#4
testcase in a separate enhancement bugreport (and yeah, -fprofile-generate
added CFG counters can twist costs of optimizations).