public inbox for gdb-patches@sourceware.org
 help / color / mirror / Atom feed
From: Pedro Alves <palves@redhat.com>
To: David Edelsohn <dje.gcc@gmail.com>
Cc: GDB Patches <gdb-patches@sourceware.org>
Subject: Re: GDB AIX build broken
Date: Tue, 25 Oct 2016 01:10:00 -0000	[thread overview]
Message-ID: <0c1a4aa3-438b-cc85-7e19-79e5912d2c37@redhat.com> (raw)
In-Reply-To: <CAGWvny=Rao6JtFbT2fNJQyGVirhB2rOCzCV6TKBJfCYmjo2Avg@mail.gmail.com>

On 10/25/2016 01:51 AM, David Edelsohn wrote:
> On Mon, Oct 24, 2016 at 8:31 PM, Pedro Alves <palves@redhat.com> wrote:
>> On 10/25/2016 01:13 AM, David Edelsohn wrote:
>>> On Mon, Oct 24, 2016 at 7:00 PM, Pedro Alves <palves@redhat.com> wrote:
>>>
>>>> That's a hint, but it can't be the fix.  common-defs.h must be the
>>>> first file included.  I suspect that gnulib's inttypes.h
>>>> replacement logic is broken on AIX.
>>>
>>> The gnulib import definitely is the commit that caused the breakage.
> 
> gnulib weaves between its header files and the system header files in
> dangerous ways.  It incorrectly assumes that it's version of a header
> file always will be seen before a system header file.  But a system
> header file may include another system header file directly.

Sure, and if so, gnulib will be patched to handle it somehow.

> 
> 
> AIX inttypes.h protects the printf format macros with
> 
> #if !defined(__cplusplus) || defined(__STDC_FORMAT_MACROS)

There you go.  See my other email.  This means AIX's
headers haven't been updated for C++11 yet.  C++11 overruled
C99's decision to require the __STDC_FORMAT_MACROS input macro,
and the C11 followed suit.

> 
> gnulib inttypes.h helpfully provides
> 
>    /* Some pre-C++11 <stdint.h> implementations need this.  */
> #  if defined __cplusplus && ! defined __STDC_FORMAT_MACROS
> #   define __STDC_FORMAT_MACROS 1
> #  endif
> 
> #  include_next <inttypes.h>
> # endif
> #endif

Right, that old way of defining __STDC_FORMAT_MACROS isn't complete.
That's why I wrote the fix at:
https://lists.gnu.org/archive/html/bug-gnulib/2015-11/msg00010.html
That now defines __STDC_FORMAT_MACROS in config.h instead.  Now the
problem is that gnulib's configure runs with a C compiler,
not a C++ compiler, so my fix doesn't really make a difference.   I had
written that assuming we'd change to configure/build gnulib
with a C++ compiler, back then.

So the fix is to do what I had suggested here
<https://sourceware.org/ml/gdb-patches/2016-10/msg00522.html>:

~~~~~~
                                              If this causes a problem
    somewhere, we can re-define the macros higher up in the file, before
    system headers are included.
~~~~~~

> Should __STDC_FORMAT_MACROS be defined in gdb/common/common-defs.h and
> gdb/defs.h (if btrace.c also is changed)?

Exactly, almost.  Only common/common-defs.h should be changed.
defs.h includes common/common-defs.h first thing,
before system headers.  ("common" refers to the file being
shared between gdb and gdbserver.)

Thanks,
Pedro Alves

      reply	other threads:[~2016-10-25  1:10 UTC|newest]

Thread overview: 16+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2016-10-24 22:40 David Edelsohn
2016-10-24 22:46 ` Pedro Alves
2016-10-24 22:53   ` David Edelsohn
2016-10-24 23:00     ` Pedro Alves
2016-10-25  0:13       ` David Edelsohn
2016-10-25  0:31         ` Pedro Alves
2016-10-25  0:50           ` Pedro Alves
2016-10-25  1:08             ` David Edelsohn
2016-10-25  1:22               ` Pedro Alves
2016-10-25 10:58                 ` [pushed] Define __STDC_CONSTANT_MACROS/__STDC_LIMIT_MACROS again (Re: GDB AIX build broken) Pedro Alves
2016-10-25 12:41                   ` David Edelsohn
2016-10-25 12:48                     ` Pedro Alves
2016-10-25 12:54                       ` [pushed] common/common-defs.h: Define __STDC_FORMAT_MACROS as well " Pedro Alves
2016-10-25 13:01                         ` David Edelsohn
2016-10-25  0:51           ` GDB AIX build broken David Edelsohn
2016-10-25  1:10             ` Pedro Alves [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=0c1a4aa3-438b-cc85-7e19-79e5912d2c37@redhat.com \
    --to=palves@redhat.com \
    --cc=dje.gcc@gmail.com \
    --cc=gdb-patches@sourceware.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).