[Bug bootstrap/98590] New: [11 regression] Bootstrap failure with Ada on Cygwin since switch to C++11

[Bug bootstrap/98590] New: [11 regression] Bootstrap failure with Ada on Cygwin since switch to C++11

[mikpelinux at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98590

            Bug ID: 98590
           Summary: [11 regression] Bootstrap failure with Ada on Cygwin
                    since switch to C++11
           Product: gcc
           Version: 11.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: bootstrap
          Assignee: unassigned at gcc dot gnu.org
          Reporter: mikpelinux at gmail dot com
  Target Milestone: ---

Attempting to bootstrap gcc-11-20210103 with Ada on x86_64-pc-cygwin fails
with:

g++ -std=c++11  -fno-PIE -c  -DIN_GCC_FRONTEND -g -DIN_GCC     -fno-exceptions
-fno-rtti -fasynchronous-unwind-tables -W -Wall -Wno-narrowing -Wwrite-strings
-Wcast-qual -Wno-error=format-diag -Wno-format  -Wmissing-format-attribute
-Woverloaded-virtual -Wno-long-long -Wno-variadic-macros
-Wno-overlength-strings   -DHAVE_CONFIG_H -I. -Iada -I/tmp/gcc-11-20210103/gcc
-I/tmp/gcc-11-20210103/gcc/ada -I/tmp/gcc-11-20210103/gcc/../include
-I/tmp/gcc-11-20210103/gcc/../libcpp/include
-I/tmp/gcc-11-20210103/gcc/../libcody
-I/home/mikpe/pkgs/cygwin-x86_64/gmp-6.2.1/include
-I/home/mikpe/pkgs/cygwin-x86_64/mpfr-4.1.0/include
-I/home/mikpe/pkgs/cygwin-x86_64/mpc-1.2.1/include 
-I/tmp/gcc-11-20210103/gcc/../libdecnumber
-I/tmp/gcc-11-20210103/gcc/../libdecnumber/bid -I../libdecnumber
-I/tmp/gcc-11-20210103/gcc/../libbacktrace   -o ada/cstreams.o -MT
ada/cstreams.o -MMD -MP -MF ada/.deps/cstreams.TPo
/tmp/gcc-11-20210103/gcc/ada/cstreams.c
In file included from /tmp/gcc-11-20210103/gcc/ada/cstreams.c:59:
/tmp/gcc-11-20210103/gcc/ada/cstreams.c: In function 'int
__gnat_fileno(FILE*)':
/tmp/gcc-11-20210103/gcc/system.h:122:26: error: 'fileno_unlocked' was not
declared in this scope; did you mean 'fopen_unlocked'?
  122 | #  define fileno(Stream) fileno_unlocked (Stream)
      |                          ^~~~~~~~~~~~~~~
/tmp/gcc-11-20210103/gcc/ada/cstreams.c:117:12: note: in expansion of macro
'fileno'
  117 |    return (fileno (stream));
      |            ^~~~~~
make[3]: *** [Makefile:1131: ada/cstreams.o] Error 1
make[3]: Leaving directory '/tmp/objdir/gcc'
make[2]: *** [Makefile:4778: all-stage1-gcc] Error 2
make[2]: Leaving directory '/tmp/objdir'
make[1]: *** [Makefile:22181: stage1-bubble] Error 2
make[1]: Leaving directory '/tmp/objdir'
make: *** [Makefile:22518: bootstrap] Error 2

GCC 7/8/9/10 all bootstrap fine with Ada (if a pending patch for PR94918 is
applied). A git bisect identified that this error started when GCC 11 switched
to requiring C++11 during bootstrap:
# first bad commit: [5329b59a2e13dabbe2038af0fe2e3cf5fc7f98ed] bootstrap:
Update requirement to C++11.

Cygwin's <stdio.h> only declares fileno_unlocked() if _GNU_SOURCE or similar is
defined. With g++ -std=c++11 or gcc -std=c11 the function is _not_ declared,
but if the -std= option is omitted or changed to a GNU dialect then the
function is declared.

For some reason however the configure checks for fileno_unlocked all succeeded:

> fgrep fileno_unlocked build.log
checking for fileno_unlocked... yes
checking whether fileno_unlocked is declared... yes
checking for fileno_unlocked... yes
checking whether fileno_unlocked is declared... yes
checking for fileno_unlocked... yes
checking whether fileno_unlocked is declared... yes
checking for fileno_unlocked... yes
checking whether fileno_unlocked is declared... yes

and left the generated config.h and auto-host.h files declaring it:

> grep -F -r HAVE_FILENO_UNLOCKED --include="*.h" .
./build-x86_64-pc-cygwin/fixincludes/config.h:#define HAVE_FILENO_UNLOCKED 1
./build-x86_64-pc-cygwin/libcpp/config.h:#define HAVE_FILENO_UNLOCKED 1
./gcc/auto-host.h:#define HAVE_FILENO_UNLOCKED 1
./libcpp/config.h:#define HAVE_FILENO_UNLOCKED 1

Seems there's a discrepancy between the configure tests and what GCC proper
does, leading to a reference to an undefined symbol.

[Bug bootstrap/98590] [11 regression] Bootstrap failure with Ada on Cygwin since switch to C++11

[mikpelinux at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98590

--- Comment #1 from Mikael Pettersson <mikpelinux at gmail dot com> ---
Created attachment 49914
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=49914&action=edit
preprocessed source for failing cstreams.c

[Bug bootstrap/98590] [11 regression] Bootstrap failure with Ada on Cygwin since switch to C++11

[mikpelinux at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98590

--- Comment #2 from Mikael Pettersson <mikpelinux at gmail dot com> ---
Created attachment 49915
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=49915&action=edit
preprocessed output with -dM -E

[Bug bootstrap/98590] [11 regression] Bootstrap failure with Ada on Cygwin since switch to C++11

[ebotcazou at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98590

Eric Botcazou <ebotcazou at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2021-01-07
                 CC|                            |ebotcazou at gcc dot gnu.org
     Ever confirmed|0                           |1

--- Comment #3 from Eric Botcazou <ebotcazou at gcc dot gnu.org> ---
It's deja vu, the default should be -std=gnu++11 instead of -std=c++11.

[Bug bootstrap/98590] [11 regression] Bootstrap failure with Ada on Cygwin since switch to C++11

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98590

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |11.0

--- Comment #4 from Richard Biener <rguenth at gcc dot gnu.org> ---
But we do not require a GNU C++11 host compiler but a C++11 host compiler so
at least for stage1 (as the description depicts) -std=c++11 is what we want
(to fend off GNU dialect usage).

So can we fix the configury discrepancy instead?

[Bug bootstrap/98590] [11 regression] Bootstrap failure with Ada on Cygwin since switch to C++11

[ebotcazou at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98590

--- Comment #5 from Eric Botcazou <ebotcazou at gcc dot gnu.org> ---
> But we do not require a GNU C++11 host compiler but a C++11 host compiler so
> at least for stage1 (as the description depicts) -std=c++11 is what we want
> (to fend off GNU dialect usage).

IMO that's ill advised and will lead to strange issues like this that should
not be worth the time of anyone.  The manual reads:

 The default, if no C++ language dialect options are given, is
'-std=gnu++xx'.

so I really don't see why we are shooting ourselves in the foot here.

[Bug bootstrap/98590] [11 regression] Bootstrap failure with Ada on Cygwin since switch to C++11

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98590

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> ---
I'm here with Richard, by using -std=c++11 for stage1 we get (at least to some
extent) verification that we aren't relying on GNU extensions and can use other
host compilers.  Of course, further stages should just use whatever g++
defaults to.

[Bug bootstrap/98590] [11 regression] Bootstrap failure with Ada on Cygwin since switch to C++11

[ebotcazou at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98590

--- Comment #7 from Eric Botcazou <ebotcazou at gcc dot gnu.org> ---
> I'm here with Richard, by using -std=c++11 for stage1 we get (at least to
> some extent) verification that we aren't relying on GNU extensions and can
> use other host compilers.  Of course, further stages should just use
> whatever g++ defaults to.

In my experience this particular extent is very limited.  If you want to use
other host compilers, then regularly test them, but do not penalize the GNU
project because of them.

[Bug bootstrap/98590] [11 regression] Bootstrap failure with Ada on Cygwin since switch to C++11

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98590

--- Comment #8 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
It is not really a penalization, the -std= is (supposed to be) added only
during bootstraps for stage1, so if one builds a cross-compiler or
non-bootstrapped compiler, it isn't added and nobody is penalized, if one
bootstraps the compiler, the only penalization will be that stage1 could be
slightly slower (it will be slow anyway, as it is built with -O0, isn't it).

[Bug bootstrap/98590] [11 regression] Bootstrap failure with Ada on Cygwin since switch to C++11

[mikpelinux at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98590

--- Comment #9 from Mikael Pettersson <mikpelinux at gmail dot com> ---
Created attachment 49925
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=49925&action=edit
ensure fileno_unlocked() is visible on Cygwin

Kludge to work around configure breakage. Not pretty, but at least it lets
bootstrap w/ Ada succeed on Cygwin.

[Bug bootstrap/98590] [11 regression] Bootstrap failure with Ada on Cygwin since switch to C++11

[ebotcazou at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98590

--- Comment #10 from Eric Botcazou <ebotcazou at gcc dot gnu.org> ---
> Kludge to work around configure breakage. Not pretty, but at least it lets
> bootstrap w/ Ada succeed on Cygwin.

OK, let's go for this.  Can you post it on the gcc-patches@ mailing list?

[Bug bootstrap/98590] [11 regression] Bootstrap failure with Ada on Cygwin since switch to C++11

[mikpelinux at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98590

--- Comment #11 from Mikael Pettersson <mikpelinux at gmail dot com> ---
(In reply to Eric Botcazou from comment #10)
> > Kludge to work around configure breakage. Not pretty, but at least it lets
> > bootstrap w/ Ada succeed on Cygwin.
> 
> OK, let's go for this.  Can you post it on the gcc-patches@ mailing list?

Done: https://gcc.gnu.org/pipermail/gcc-patches/2021-February/564804.html

Sorry about the delay, I had issues building current trunk on Cygwin, with
symptoms similar to those shown in the DWARF5 regression PRs.

[Bug bootstrap/98590] [11 regression] Bootstrap failure with Ada on Cygwin since switch to C++11

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98590

--- Comment #12 from Richard Biener <rguenth at gcc dot gnu.org> ---
Looks like the fix was approved here
https://gcc.gnu.org/pipermail/gcc-patches/2021-February/564929.html but not yet
committed.

[Bug bootstrap/98590] [11 regression] Bootstrap failure with Ada on Cygwin since switch to C++11

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98590

--- Comment #13 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Jeff Law <law@gcc.gnu.org>:

https://gcc.gnu.org/g:8b6ebc025cf2b25fdc1e8f6e6261701dc71bac74

commit r11-7464-g8b6ebc025cf2b25fdc1e8f6e6261701dc71bac74
Author: Mikael Pettersson <mikpelinux@gmail.com>
Date:   Tue Mar 2 15:20:06 2021 -0700

    [PATCH] Fix Ada bootstrap failure on Cygwin since switch to C++11 (PR98590)

    gcc/ada
            PR bootstrap/98590
            * cstreams.c: Ensure fileno_unlocked() is visible on Cygwin.

[Bug bootstrap/98590] [11 regression] Bootstrap failure with Ada on Cygwin since switch to C++11

[mikpelinux at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98590

Mikael Pettersson <mikpelinux at gmail dot com> changed:

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

--- Comment #14 from Mikael Pettersson <mikpelinux at gmail dot com> ---
This issue is now fixed.
Bootstrap still needs the PR94918 fix though.