wrong bfd recognized

wrong bfd recognized

[Luca Pizzamiglio]

[-- Attachment #1: Type: text/plain, Size: 553 bytes --]

Hi!
I'm Luca, the maintainer of gdb7.x on FreeBSD ports.

I've found a strange situation where gdb's configure script fails to
use libbfd and disable ELF support.
the problem was that during configuration, the /usr/local/lib/libbfd.a
was used, instead of the gdb's one. This library was compiled with
libintl support, then the bfd's test program build failed.

This patch should solve this issue, building bfd's testing program
using the gdb's one.

Thanks in advance for the support

Luca

PS: please use my address in cc, I'm not in the mailing-list!

[-- Attachment #2: gdb-configure.ac.patch --]
[-- Type: application/octet-stream, Size: 479 bytes --]

--- gdb/configure.ac.orig	2011-12-13 15:33:27.616649109 +0100
+++ gdb/configure.ac	2011-12-13 15:34:08.559285673 +0100
@@ -1854,7 +1854,7 @@
 OLD_LDFLAGS=$LDFLAGS
 OLD_LIBS=$LIBS
 CFLAGS="$CFLAGS -I${srcdir}/../include -I../bfd -I${srcdir}/../bfd"
-LDFLAGS="$LDFLAGS -L../bfd -L../libiberty"
+LDFLAGS="-L../bfd -L../libiberty $LDFLAGS"
 intl=`echo $LIBINTL | sed 's,${top_builddir}/,,g'`
 # -ldl is provided by bfd/Makfile.am (LIBDL) <PLUGINS>.
 if test "$plugins" = "yes"; then

Re: wrong bfd recognized

[Tom Tromey]

>>>>> "Luca" == Luca Pizzamiglio <luca.pizzamiglio@gmail.com> writes:

Luca> I've found a strange situation where gdb's configure script fails to
Luca> use libbfd and disable ELF support.
Luca> the problem was that during configuration, the /usr/local/lib/libbfd.a
Luca> was used, instead of the gdb's one. This library was compiled with
Luca> libintl support, then the bfd's test program build failed.

Are you adding -L/usr/local/lib to LDFLAGS?
If so -- why?

Luca> This patch should solve this issue, building bfd's testing program
Luca> using the gdb's one.

I wonder whether it would break something.

If you can find another spot in the src configury that does it this way,
I guess I'd be more inclined to just do it.

This needs a ChangeLog entry.

Tom

Re: wrong bfd recognized

[Luca Pizzamiglio]

Hi Tom,

After a check, I've noticed that LDFLAGS is set -L/usr/local/lib in
order to use libreadline installed there.
It's possible to have several libreadline installed and LDFLAGS is the
only way to specify which library to use.
On details. FreeBSD could have a libreadline provided by the system
and a libreadline provided by ports (different versions, license,
ecc.)
So, if LDFLAGS is set, then the gdb's libbfd is not used, but the
/usr/local/lib one's.

I'm not totally sure, but I guess that this behavior is not properly correct.
Am I missing some contraindication or a way to point the right libreadline?

thanks in advance for the help

Luca

On Tue, Dec 20, 2011 at 3:56 PM, Tom Tromey <tromey@redhat.com> wrote:
>>>>>> "Luca" == Luca Pizzamiglio <luca.pizzamiglio@gmail.com> writes:
>
> Luca> I've found a strange situation where gdb's configure script fails to
> Luca> use libbfd and disable ELF support.
> Luca> the problem was that during configuration, the /usr/local/lib/libbfd.a
> Luca> was used, instead of the gdb's one. This library was compiled with
> Luca> libintl support, then the bfd's test program build failed.
>
> Are you adding -L/usr/local/lib to LDFLAGS?
> If so -- why?
>
> Luca> This patch should solve this issue, building bfd's testing program
> Luca> using the gdb's one.
>
> I wonder whether it would break something.
>
> If you can find another spot in the src configury that does it this way,
> I guess I'd be more inclined to just do it.
>
> This needs a ChangeLog entry.
>
> Tom

Re: wrong bfd recognized

[Tom Tromey]

>>>>> "Luca" == Luca Pizzamiglio <luca.pizzamiglio@gmail.com> writes:

Luca> Am I missing some contraindication or a way to point the right
Luca> libreadline?

I don't think there is.

Luca> thanks in advance for the help

Could you write a ChangeLog entry for your patch?
I will put it in.

Tom

Re: wrong bfd recognized

[Luca Pizzamiglio]

Ok. A possible entry could be:

improved bfd local static library detection in the configure script

Luca

PS Sorry for the delay, Xmas break ;)



On Wed, Dec 21, 2011 at 6:28 PM, Tom Tromey <tromey@redhat.com> wrote:
>>>>>> "Luca" == Luca Pizzamiglio <luca.pizzamiglio@gmail.com> writes:
>
> Luca> Am I missing some contraindication or a way to point the right
> Luca> libreadline?
>
> I don't think there is.
>
> Luca> thanks in advance for the help
>
> Could you write a ChangeLog entry for your patch?
> I will put it in.
>
> Tom

Re: wrong bfd recognized

[Joel Brobecker]

Luca

> > Could you write a ChangeLog entry for your patch?
> > I will put it in.
[...]
> Ok. A possible entry could be:
> 
> improved bfd local static library detection in the configure script

Unfortunately, this is not what a ChangeLog entry should look like.
I suspect you thought that we were asking for some text to be used
as revision history?  Here is what a ChangeLog entry looks like:

2011-12-27  Doug Evans  <dje@google.com>

        * dwarf2read.c (struct dwarf2_cu): Delete members first_fn, last_fn,
        cached_fn.
        (struct function_range): Delete.
        (initialize_cu_func_list, add_to_cu_func_list): Delete.  All callers
        updated.
        (check_cu_functions): Ditto.

You can have a look at the couple of sections that explain ChangeLogs
in the GNU Coding Standards (the link brings you directly to the
relevant section):

    http://www.gnu.org/prep/standards/standards.html#Change-Logs

Please let me know if you have any problems creating one and I will
write it for you (this time!).

-- 
Joel

Re: wrong bfd recognized

[Luca Pizzamiglio]

Hi Joel,

sorry for the confusion, I'm not used to write this kind of ChangeLog
entry. My fault.

2012-01-02 Luca Pizzamiglio <luca.pizzamiglio@gmail.com>

    * gdb/configure.ac (LDFLAGS in ELF support's check): changing path
priority in LDFLAGS
    LDFLAGS is redefined putting an external-defined value at the end.

I hope that works now! This rules doesn't fit well with autoconf files :)
Feel free to modify it, if it's still wrong.

thanks for the support

Luca

On Thu, Dec 29, 2011 at 1:17 PM, Joel Brobecker <brobecker@adacore.com> wrote:
> Luca
>
>> > Could you write a ChangeLog entry for your patch?
>> > I will put it in.
> [...]
>> Ok. A possible entry could be:
>>
>> improved bfd local static library detection in the configure script
>
> Unfortunately, this is not what a ChangeLog entry should look like.
> I suspect you thought that we were asking for some text to be used
> as revision history?  Here is what a ChangeLog entry looks like:
>
> 2011-12-27  Doug Evans  <dje@google.com>
>
>        * dwarf2read.c (struct dwarf2_cu): Delete members first_fn, last_fn,
>        cached_fn.
>        (struct function_range): Delete.
>        (initialize_cu_func_list, add_to_cu_func_list): Delete.  All callers
>        updated.
>        (check_cu_functions): Ditto.
>
> You can have a look at the couple of sections that explain ChangeLogs
> in the GNU Coding Standards (the link brings you directly to the
> relevant section):
>
>    http://www.gnu.org/prep/standards/standards.html#Change-Logs
>
> Please let me know if you have any problems creating one and I will
> write it for you (this time!).
>
> --
> Joel

Re: wrong bfd recognized

[Tom Tromey]

>>>>> "Luca" == Luca Pizzamiglio <luca.pizzamiglio@gmail.com> writes:

Luca> sorry for the confusion, I'm not used to write this kind of ChangeLog
Luca> entry. My fault.

No worries.

I'm sorry about the delay on this.  I'm finally checking it in, with
this ChangeLog entry:

2012-02-06  Luca Pizzamiglio  <luca.pizzamiglio@gmail.com>

	* configure: Rebuild.
	* configure.ac: Put -L../bfd and -L../libiberty at the front of
	LDFLAGS.

Tom

Re: wrong bfd recognized

[Luca Pizzamiglio]

Thanks Tom!

Best regards,
Luca

On Mon, Feb 6, 2012 at 8:32 PM, Tom Tromey <tromey@redhat.com> wrote:
>>>>>> "Luca" == Luca Pizzamiglio <luca.pizzamiglio@gmail.com> writes:
>
> Luca> sorry for the confusion, I'm not used to write this kind of ChangeLog
> Luca> entry. My fault.
>
> No worries.
>
> I'm sorry about the delay on this.  I'm finally checking it in, with
> this ChangeLog entry:
>
> 2012-02-06  Luca Pizzamiglio  <luca.pizzamiglio@gmail.com>
>
>        * configure: Rebuild.
>        * configure.ac: Put -L../bfd and -L../libiberty at the front of
>        LDFLAGS.
>
> Tom

Re: wrong bfd recognized

[Pedro Alves]

On 12/13/2011 03:13 PM, Luca Pizzamiglio wrote:
> +++ gdb/configure.ac	2011-12-13 15:34:08.559285673 +0100
> @@ -1854,7 +1854,7 @@
>  OLD_LDFLAGS=$LDFLAGS
>  OLD_LIBS=$LIBS
>  CFLAGS="$CFLAGS -I${srcdir}/../include -I../bfd -I${srcdir}/../bfd"
> -LDFLAGS="$LDFLAGS -L../bfd -L../libiberty"
> +LDFLAGS="-L../bfd -L../libiberty $LDFLAGS"

If you set LDFLAGS to -L/foo/lib , you presumably also set
CFLAGS to -I/foo/include .  In that case, after the patch, if there's a libbfd
in /foo/, you'll run the test with gdb's bfd, but compile it against e.g., the
bfd.h from /foo/include.  You may not see this happen with /usr/local/include
because gcc usually ignores -I/usr/local/include as being a duplicate of a
system header path.

IOW, it'd make sense to also do:

- CFLAGS="$CFLAGS -I${srcdir}/../include -I../bfd -I${srcdir}/../bfd"
+ CFLAGS="-I${srcdir}/../include -I../bfd -I${srcdir}/../bfd $CFLAGS"

-- 
Pedro Alves

Re: wrong bfd recognized

[Pedro Alves]

On 02/07/2012 08:15 PM, Pedro Alves wrote:
> On 12/13/2011 03:13 PM, Luca Pizzamiglio wrote:
>> +++ gdb/configure.ac	2011-12-13 15:34:08.559285673 +0100
>> @@ -1854,7 +1854,7 @@
>>  OLD_LDFLAGS=$LDFLAGS
>>  OLD_LIBS=$LIBS
>>  CFLAGS="$CFLAGS -I${srcdir}/../include -I../bfd -I${srcdir}/../bfd"
>> -LDFLAGS="$LDFLAGS -L../bfd -L../libiberty"
>> +LDFLAGS="-L../bfd -L../libiberty $LDFLAGS"
> 
> If you set LDFLAGS to -L/foo/lib , you presumably also set
> CFLAGS to -I/foo/include .  In that case, after the patch, if there's a libbfd
> in /foo/, you'll run the test with gdb's bfd, but compile it against e.g., the
> bfd.h from /foo/include.  You may not see this happen with /usr/local/include
> because gcc usually ignores -I/usr/local/include as being a duplicate of a
> system header path.
> 
> IOW, it'd make sense to also do:
> 
> - CFLAGS="$CFLAGS -I${srcdir}/../include -I../bfd -I${srcdir}/../bfd"
> + CFLAGS="-I${srcdir}/../include -I../bfd -I${srcdir}/../bfd $CFLAGS"
> 

I'm applying this.

2012-02-10  Pedro Alves  <palves@redhat.com>

	* configure.ac (HAVE_ELF test): Put the old CFLAGS at the end of
	the test CFLAGS.
	* configure: Regenerate.

---

 gdb/configure    |    5 ++++-
 gdb/configure.ac |    5 ++++-
 2 files changed, 8 insertions(+), 2 deletions(-)

diff --git a/gdb/configure b/gdb/configure
index 11c044c..2566410 100755
--- a/gdb/configure
+++ b/gdb/configure
@@ -15320,7 +15320,10 @@ esac
 OLD_CFLAGS=$CFLAGS
 OLD_LDFLAGS=$LDFLAGS
 OLD_LIBS=$LIBS
-CFLAGS="$CFLAGS -I${srcdir}/../include -I../bfd -I${srcdir}/../bfd"
+# Put the old CFLAGS/LDFLAGS last, in case the user's (C|LD)FLAGS
+# points somewhere with bfd, with -I/foo/lib and -L/foo/lib.  We
+# always want our bfd.
+CFLAGS="-I${srcdir}/../include -I../bfd -I${srcdir}/../bfd $CFLAGS"
 LDFLAGS="-L../bfd -L../libiberty $LDFLAGS"
 intl=`echo $LIBINTL | sed 's,${top_builddir}/,,g'`
 # -ldl is provided by bfd/Makfile.am (LIBDL) <PLUGINS>.
diff --git a/gdb/configure.ac b/gdb/configure.ac
index 36da463..1b11adb 100644
--- a/gdb/configure.ac
+++ b/gdb/configure.ac
@@ -1908,7 +1908,10 @@ AC_SUBST(WIN32LIBS)
 OLD_CFLAGS=$CFLAGS
 OLD_LDFLAGS=$LDFLAGS
 OLD_LIBS=$LIBS
-CFLAGS="$CFLAGS -I${srcdir}/../include -I../bfd -I${srcdir}/../bfd"
+# Put the old CFLAGS/LDFLAGS last, in case the user's (C|LD)FLAGS
+# points somewhere with bfd, with -I/foo/lib and -L/foo/lib.  We
+# always want our bfd.
+CFLAGS="-I${srcdir}/../include -I../bfd -I${srcdir}/../bfd $CFLAGS"
 LDFLAGS="-L../bfd -L../libiberty $LDFLAGS"
 intl=`echo $LIBINTL | sed 's,${top_builddir}/,,g'`
 # -ldl is provided by bfd/Makfile.am (LIBDL) <PLUGINS>.

Re: wrong bfd recognized

[Luca Pizzamiglio]

Hi Pedro,

sorry for the delay.
You're right. also CFLAG should be updated.
I didn't see it, because it has generated no problem.

Thanks

Best regards,
Luca


On Fri, Feb 10, 2012 at 2:54 PM, Pedro Alves <palves@redhat.com> wrote:
> On 02/07/2012 08:15 PM, Pedro Alves wrote:
>> On 12/13/2011 03:13 PM, Luca Pizzamiglio wrote:
>>> +++ gdb/configure.ac 2011-12-13 15:34:08.559285673 +0100
>>> @@ -1854,7 +1854,7 @@
>>>  OLD_LDFLAGS=$LDFLAGS
>>>  OLD_LIBS=$LIBS
>>>  CFLAGS="$CFLAGS -I${srcdir}/../include -I../bfd -I${srcdir}/../bfd"
>>> -LDFLAGS="$LDFLAGS -L../bfd -L../libiberty"
>>> +LDFLAGS="-L../bfd -L../libiberty $LDFLAGS"
>>
>> If you set LDFLAGS to -L/foo/lib , you presumably also set
>> CFLAGS to -I/foo/include .  In that case, after the patch, if there's a libbfd
>> in /foo/, you'll run the test with gdb's bfd, but compile it against e.g., the
>> bfd.h from /foo/include.  You may not see this happen with /usr/local/include
>> because gcc usually ignores -I/usr/local/include as being a duplicate of a
>> system header path.
>>
>> IOW, it'd make sense to also do:
>>
>> - CFLAGS="$CFLAGS -I${srcdir}/../include -I../bfd -I${srcdir}/../bfd"
>> + CFLAGS="-I${srcdir}/../include -I../bfd -I${srcdir}/../bfd $CFLAGS"
>>
>
> I'm applying this.
>
> 2012-02-10  Pedro Alves  <palves@redhat.com>
>
>        * configure.ac (HAVE_ELF test): Put the old CFLAGS at the end of
>        the test CFLAGS.
>        * configure: Regenerate.
>
> ---
>
>  gdb/configure    |    5 ++++-
>  gdb/configure.ac |    5 ++++-
>  2 files changed, 8 insertions(+), 2 deletions(-)
>
> diff --git a/gdb/configure b/gdb/configure
> index 11c044c..2566410 100755
> --- a/gdb/configure
> +++ b/gdb/configure
> @@ -15320,7 +15320,10 @@ esac
>  OLD_CFLAGS=$CFLAGS
>  OLD_LDFLAGS=$LDFLAGS
>  OLD_LIBS=$LIBS
> -CFLAGS="$CFLAGS -I${srcdir}/../include -I../bfd -I${srcdir}/../bfd"
> +# Put the old CFLAGS/LDFLAGS last, in case the user's (C|LD)FLAGS
> +# points somewhere with bfd, with -I/foo/lib and -L/foo/lib.  We
> +# always want our bfd.
> +CFLAGS="-I${srcdir}/../include -I../bfd -I${srcdir}/../bfd $CFLAGS"
>  LDFLAGS="-L../bfd -L../libiberty $LDFLAGS"
>  intl=`echo $LIBINTL | sed 's,${top_builddir}/,,g'`
>  # -ldl is provided by bfd/Makfile.am (LIBDL) <PLUGINS>.
> diff --git a/gdb/configure.ac b/gdb/configure.ac
> index 36da463..1b11adb 100644
> --- a/gdb/configure.ac
> +++ b/gdb/configure.ac
> @@ -1908,7 +1908,10 @@ AC_SUBST(WIN32LIBS)
>  OLD_CFLAGS=$CFLAGS
>  OLD_LDFLAGS=$LDFLAGS
>  OLD_LIBS=$LIBS
> -CFLAGS="$CFLAGS -I${srcdir}/../include -I../bfd -I${srcdir}/../bfd"
> +# Put the old CFLAGS/LDFLAGS last, in case the user's (C|LD)FLAGS
> +# points somewhere with bfd, with -I/foo/lib and -L/foo/lib.  We
> +# always want our bfd.
> +CFLAGS="-I${srcdir}/../include -I../bfd -I${srcdir}/../bfd $CFLAGS"
>  LDFLAGS="-L../bfd -L../libiberty $LDFLAGS"
>  intl=`echo $LIBINTL | sed 's,${top_builddir}/,,g'`
>  # -ldl is provided by bfd/Makfile.am (LIBDL) <PLUGINS>.

Re: wrong bfd recognized

[Pedro Alves]

On 02/10/2012 01:58 PM, Luca Pizzamiglio wrote:
> Hi Pedro,
> 
> sorry for the delay.

No worries.

> You're right. also CFLAG should be updated.
> I didn't see it, because it has generated no problem.

Yeah, because -I/usr/local/include gets special treatment
by the compiler (it's ignored).  Something like, say, -I/opt/tools/include
would cause you trouble if you had put both readline and bfd there.

Thanks,
-- 
Pedro Alves