[Bug middle-end/64340] New: [5 Regression] FAIL: gnat.dg/lto8.adb (internal compiler error)

[Bug middle-end/64340] New: [5 Regression] FAIL: gnat.dg/lto8.adb (internal compiler error)

[dominiq at lps dot ens.fr]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64340

            Bug ID: 64340
           Summary: [5 Regression] FAIL: gnat.dg/lto8.adb (internal
                    compiler error)
           Product: gcc
           Version: 5.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: middle-end
          Assignee: unassigned at gcc dot gnu.org
          Reporter: dominiq at lps dot ens.fr

Between revisions r218766 (OK) and r218797 (ICE) the test gnat.dg/lto8.adb has
started to fail. The ICE is

/opt/gcc/p_work/gcc/testsuite/gnat.dg/lto8_pkg.ads: In function
'lto8_pkg___elabs':
/opt/gcc/p_work/gcc/testsuite/gnat.dg/lto8_pkg.ads:4:1: internal compiler
error: in expand_gimple_stmt_1, at cfgexpand.c:3420

[Bug middle-end/64340] [5 Regression] FAIL: gnat.dg/lto8.adb (internal compiler error)

[schwab@linux-m68k.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64340

--- Comment #1 from Andreas Schwab <schwab@linux-m68k.org> ---
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64043#c11

[Bug middle-end/64340] [5 Regression] FAIL: gnat.dg/lto8.adb (internal compiler error)

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64340

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |5.0

[Bug middle-end/64340] [5 Regression] FAIL: gnat.dg/lto8.adb (internal compiler error)

[hubicka at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64340

Jan Hubicka <hubicka at gcc dot gnu.org> changed:

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

--- Comment #2 from Jan Hubicka <hubicka at gcc dot gnu.org> ---
Seems more of preloading needs tobe skipped.  I will dig into that.

[Bug middle-end/64340] [5 Regression] FAIL: gnat.dg/lto8.adb (internal compiler error)

[danglin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64340

John David Anglin <danglin at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2014-12-17
                 CC|                            |danglin at gcc dot gnu.org
     Ever confirmed|0                           |1

--- Comment #3 from John David Anglin <danglin at gcc dot gnu.org> ---
Also seen on hppa-unknown-linux-gnu.

[Bug middle-end/64340] [5 Regression] FAIL: gnat.dg/lto8.adb (internal compiler error)

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64340

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P1

[Bug middle-end/64340] [5 Regression] FAIL: gnat.dg/lto8.adb (internal compiler error)

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64340

--- Comment #4 from Richard Biener <rguenth at gcc dot gnu.org> ---
ISTR that preloading pid_t was necessary for ada (basically all types required
from some builtins are required to be preloaded).

[Bug middle-end/64340] [5 Regression] FAIL: gnat.dg/lto8.adb (internal compiler error)

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64340

--- Comment #5 from Richard Biener <rguenth at gcc dot gnu.org> ---
It's fixed by dropping all of TI_TARGET_OPTION_DEFAULT,
TI_TARGET_OPTION_CURRENT, TI_CURRENT_TARGET_PRAGMA and
TI_CURRENT_OPTIMIZE_PRAGMA from the special-casing.

We are trying to expand a GIMPLE_RESX stmt which should have been lowered
before.

(gdb) p opts_for_fn (cfun->decl)->x_flag_exceptions
$3 = 0 '\000'

thus somehow -fexceptions is bogus (also in 'Optimization' for some weird
reason).

(gdb) p global_options.x_flag_exceptions
$4 = 0

But certainly -fexceptions -fnon-call-exceptions was specified for both
source TUs and thus via lto-wrapper xgcc should have received them.  But
lto1 seems to ignore -fexceptions.

I've reproduced pasting the gnatmake command and then (gnatlink fails)

./lto1 -quiet lto8.o lto8_pkg.o -fexceptions -fnon-call-exceptions

The bogus setting is "restored" via

Hardware watchpoint 5: *$3

Old value = 1
New value = 0
cl_optimization_restore (opts=0x203fc80 <global_options>, ptr=0x7ffff6c344b0)
    at options-save.c:476
476       opts->x_flag_expensive_optimizations =
ptr->x_flag_expensive_optimizations;

#0  cl_optimization_restore (opts=0x203fc80 <global_options>, 
    ptr=0x7ffff6c344b0) at options-save.c:476
#1  0x00000000008f4603 in invoke_set_current_function_hook (fndecl=
    <function_decl 0x7ffff6e1a000 lto8___finalizer>)
    at /space/rguenther/src/svn/trunk2/gcc/function.c:4665
#2  0x00000000008f4909 in allocate_struct_function (
    fndecl=<function_decl 0x7ffff6e1a000 lto8___finalizer>, abstract_p=false)
    at /space/rguenther/src/svn/trunk2/gcc/function.c:4777
#3  0x00000000008f4b09 in push_struct_function (
    fndecl=<function_decl 0x7ffff6e1a000 lto8___finalizer>)
    at /space/rguenther/src/svn/trunk2/gcc/function.c:4819
#4  0x0000000000aed772 in input_function (
    fn_decl=<function_decl 0x7ffff6e1a000 lto8___finalizer>, 
    data_in=0x2062810, ib=0x7fffffffdad0, ib_cfg=0x7fffffffdac0)
    at /space/rguenther/src/svn/trunk2/gcc/lto-streamer-in.c:945
#5  0x0000000000aee06d in lto_read_body_or_constructor (
    file_data=0x7ffff7ff6000, node=0x7ffff6c347a8, data=0x20a2540 "\004", 
    section_type=LTO_section_function_body)
    at /space/rguenther/src/svn/trunk2/gcc/lto-streamer-in.c:1108
#6  0x0000000000aee311 in lto_input_function_body (file_data=0x7ffff7ff6000, 
    node=<cgraph_node* 0x7ffff6c347a8 "lto8___finalizer">, 
    data=0x20a2540 "\004")
    at /space/rguenther/src/svn/trunk2/gcc/lto-streamer-in.c:1153

now the question is why there is false for those options in the function
specific optimization nodes.  When do we populate these?

[Bug middle-end/64340] [5 Regression] FAIL: gnat.dg/lto8.adb (internal compiler error)

[tbsaunde at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64340

tbsaunde at gcc dot gnu.org changed:

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

--- Comment #6 from tbsaunde at gcc dot gnu.org ---
It seems like the right thing to do is make varpool_node::create_alias() set
ref_list.references, does that seem correct?

[Bug middle-end/64340] [5 Regression] FAIL: gnat.dg/lto8.adb (internal compiler error)

[tbsaunde at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64340

--- Comment #7 from tbsaunde at gcc dot gnu.org ---
(In reply to tbsaunde from comment #6)
> It seems like the right thing to do is make varpool_node::create_alias() set
> ref_list.references, does that seem correct?

O.O that was for another bug.

[Bug middle-end/64340] [5 Regression] FAIL: gnat.dg/lto8.adb (internal compiler error)

[hubicka at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64340

--- Comment #8 from Jan Hubicka <hubicka at gcc dot gnu.org> ---
Richard, isn't it the streamer being out of sync again - the prevoius problem
(reason I dropped pid_t) is that at the moment we preload something that is
NULL in one FE and not in another, we end up with tree indexes being wrong.

cl_optimization_restore just restore global flags from TARGET node, for
complation with -O2 it seems bogus it would be set to 0.

[Bug middle-end/64340] [5 Regression] FAIL: gnat.dg/lto8.adb (internal compiler error)

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64340

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

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

--- Comment #9 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Created attachment 34687
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=34687&action=edit
gcc5-pr64340.patch

Untested fix (well, tested with make check-gnat).

The problem seems to be that gnat modifies global_options after toplevel.c
(process_options) ends, in this case options like flag_*exceptions*, in other
cases flag_strict_aliasing, or other flags.  That means those option changes
aren't visible in optimization_{default,current}_node that is created at the
end of process_options, and if some function doesn't use its own optimization
node, it will get the default one.  So, if there are any functions with
non-default optimization node, or when lto streams that node for functions that
didn't have any, those option changes may get lost.

Seems those option changes depend on parsing, so process_options is too early,
so IMHO we need to recreate the optimization_{default,current}_node trees with
the updated options (updating the options in there directly doesn't work, as
the options are hashed).

Eric/Honza, does this look reasonable to you?

flag_ipa_ra change seems to be misplaced too, will fix that in a separate
patch.

[Bug middle-end/64340] [5 Regression] FAIL: gnat.dg/lto8.adb (internal compiler error)

[ebotcazou at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64340

--- Comment #10 from Eric Botcazou <ebotcazou at gcc dot gnu.org> ---
> Seems those option changes depend on parsing, so process_options is too
> early, so IMHO we need to recreate the optimization_{default,current}_node
> trees with the updated options (updating the options in there directly
> doesn't work, as the options are hashed).

Yes, we cannot changes these options earlier.

> Eric/Honza, does this look reasonable to you?

Maybe "Save the current optimization options again after the above changes".

> flag_ipa_ra change seems to be misplaced too, will fix that in a separate
> patch.

Yep, this looks problematic too.  Thanks for looking into this.

[Bug middle-end/64340] [5 Regression] FAIL: gnat.dg/lto8.adb (internal compiler error)

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64340

--- Comment #11 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Author: jakub
Date: Sat Feb  7 14:44:26 2015
New Revision: 220504

URL: https://gcc.gnu.org/viewcvs?rev=220504&root=gcc&view=rev
Log:
    PR middle-end/64340
    * gcc-interface/trans.c (gigi): Recreate optimization_default_node
    and optimization_current_node after tweaking global_options.

Modified:
    trunk/gcc/ada/ChangeLog
    trunk/gcc/ada/gcc-interface/trans.c

[Bug middle-end/64340] [5 Regression] FAIL: gnat.dg/lto8.adb (internal compiler error)

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64340

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

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

--- Comment #12 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Fixed.