Building binutils 2.18.50 on Ubuntu 8.10 (gcc 4.3.2)

[jb-59]

Hi all,

I've been testing building the toolchain on the most recent major
Ubuntu release, 8.10, and it appears that binutils breaks due to
stricter enforcement of string literals being passed to things like
printf and fprintf. For instance, out of the box, binutils 2.18.50
with the appropriate patch errors out the first of two times on
binutils-2.18.50/binutils/dwarf.c like so:

gcc -DHAVE_CONFIG_H -I.
-I/opt/or1200-toolchain/build/binutils-2.18.50/binutils -I. -I.
-I/opt/or1200-toolchain/build/binutils-2.18.50/binutils -I../bfd
-I/opt/or1200-toolchain/build/binutils-2.18.50/binutils/../bfd
-I/opt/or1200-toolchain/build/binutils-2.18.50/binutils/../include
-DLOCALEDIR="\"/opt/or1200-toolchain/build/tools/or32-uclinux/share/locale\""
-Dbin_dummy_emulation=bin_vanilla_emulation   -W -Wall
-Wstrict-prototypes -Wmissing-prototypes -Werror -g -O2 -c
/opt/or1200-toolchain/build/binutils-2.18.50/binutils/dwarf.c
cc1: warnings being treated as errors
/opt/or1200-toolchain/build/binutils-2.18.50/binutils/dwarf.c: In
function ‘print_dwarf_vma’:
/opt/or1200-toolchain/build/binutils-2.18.50/binutils/dwarf.c:189:
error: format not a string literal and no format arguments
make[4]: *** [dwarf.o] Error 1


It was configured from a recent version of the MOF script, like so:
$BUILD_TOP/$BINUTILS_VER/configure --target=or32-uclinux
--prefix=$BUILD_TOP/tools/or32-uclinux --disable-checking --disable-Werror

So, the line it refers to in dwarf.c that it doesn't like is like this:
 printf (buff + (byte_size == 4 ? 8 : 0));

Now, I'm by no means one hundred percent that this is correct, but I
presume there's a pre-made string it wants to pass to printf, so if I
just put in a "%s" in there it will fix this error? Wary that I may
have changed the code so that it doesn't do what was intended, I can
at least make the file compile by changing the line at 189 to:
printf ("%s", buff + (byte_size == 4 ? 8 : 0));


The next issue when compiling is for the same reason, in the file
binutils-2.18.50/gas/listing.c:

gcc -DHAVE_CONFIG_H -I.
-I/opt/or1200-toolchain/build/binutils-2.18.50/gas -I. -I.
-I/opt/or1200-toolchain/build/binutils-2.18.50/gas -I../bfd
-I/opt/or1200-toolchain/build/binutils-2.18.50/gas/config
-I/opt/or1200-toolchain/build/binutils-2.18.50/gas/../include
-I/opt/or1200-toolchain/build/binutils-2.18.50/gas/..
-I/opt/or1200-toolchain/build/binutils-2.18.50/gas/../bfd
-DLOCALEDIR="\"/opt/or1200-toolchain/build/tools/or32-uclinux/share/locale\""
  -W -Wall -Wstrict-prototypes -Wmissing-prototypes -Werror -g -O2 -c
/opt/or1200-toolchain/build/binutils-2.18.50/gas/listing.c
cc1: warnings being treated as errors
/opt/or1200-toolchain/build/binutils-2.18.50/gas/listing.c: In
function ‘print_options’:
/opt/or1200-toolchain/build/binutils-2.18.50/gas/listing.c:1103:
error: format not a string literal and no format arguments
make[4]: *** [listing.o] Error 1


The line concerned in listing.c is:

fprintf (list_file, field_name);

So I presume here we put in a format string like before, again wary
that I have not much of an idea of what this is meant to do:
fprintf (list_file, "%s", field_name);


And now binutils-2.18.50 compiles and installs.

GCC-4.2.2 seems to compile OK too, however the next issues with this
toolchain build are to do with the linux kernel portion, but I'm yet
to track down the specifics of this. I'll post more when I know.

Anyway, i'm not sure if perhaps there's a configure option I can pass
(I've seen the -Wformat-nonliteral, but I guess this makes gcc thrown
an error at any of these warnings.... which is now standard behavior
in gcc-4.3.2 I think) or perhaps these fixes should be in the next
patch (or perhaps already are fixed in binutils-2.19). Anyone know
what could be done to make this work first time, every time?

-jb
_______________________________________________
http://www.opencores.org/mailman/listinfo/openrisc

[jb-59]

Well, I've got the answer to my own problem here.

It appears that this change in behavior of gcc-4.3.2 has been
addressed in binutils-2.19, those two lines have been changed to be,
in the case of binutils-2.19/binutils/dwarf.c:
fputs (buff + (byte_size == 4 ? 8 : 0), stdout);

and in binutils-2.19/gas/listing.c:
fputs (field_name, list_file);

fputs() saves the day from nasty string literal errors.

----- Original Message -----
From: jb at orsoc.se<jb@o...>
To:
Date: Thu Mar  5 11:25:19 CET 2009
Subject: [openrisc] Building binutils 2.18.50 on Ubuntu 8.10 (gcc
4.3.2)

> Hi all,
>
> I've been testing building the toolchain on the most recent major
> Ubuntu release, 8.10, and it appears that binutils breaks due to
> stricter enforcement of string literals being passed to things like
> printf and fprintf. For instance, out of the box, binutils 2.18.50
> with the appropriate patch errors out the first of two times on
> binutils-2.18.50/binutils/dwarf.c like so:
> gcc -DHAVE_CONFIG_H -I.
> -I/opt/or1200-toolchain/build/binutils-2.18.50/binutils -I. -I.
> -I/opt/or1200-toolchain/build/binutils-2.18.50/binutils -I../bfd
> -I/opt/or1200-toolchain/build/binutils-2.18.50/binutils/../bfd
> -I/opt/or1200-toolchain/build/binutils-2.18.50/binutils/../include
>
>

-DLOCALEDIR="\"/opt/or1200-toolchain/build/tools/or32-uclinux/share/locale\""

> -Dbin_dummy_emulation=bin_vanilla_emulation -W -Wall
> -Wstrict-prototypes -Wmissing-prototypes -Werror -g -O2 -c
> /opt/or1200-toolchain/build/binutils-2.18.50/binutils/dwarf.c
> cc1: warnings being treated as errors
> /opt/or1200-toolchain/build/binutils-2.18.50/binutils/dwarf.c: In
> function ‘print_dwarf_vma’:
> /opt/or1200-toolchain/build/binutils-2.18.50/binutils/dwarf.c:189:
> error: format not a string literal and no format arguments
> make[4]: *** [dwarf.o] Error 1
> It was configured from a recent version of the MOF script, like so:
> $BUILD_TOP/$BINUTILS_VER/configure --target=or32-uclinux
> --prefix=$BUILD_TOP/tools/or32-uclinux --disable-checking
> --disable-Werror
> So, the line it refers to in dwarf.c that it doesn't like is like
> this:
>  printf (buff + (byte_size == 4 ? 8 : 0));
> Now, I'm by no means one hundred percent that this is correct, but
> I
> presume there's a pre-made string it wants to pass to printf, so if
> I
> just put in a "%s" in there it will fix this error? Wary
> that I may
> have changed the code so that it doesn't do what was intended, I
> can
> at least make the file compile by changing the line at 189 to:
> printf ("%s", buff + (byte_size == 4 ? 8 : 0));
> The next issue when compiling is for the same reason, in the file
> binutils-2.18.50/gas/listing.c:
> gcc -DHAVE_CONFIG_H -I.
> -I/opt/or1200-toolchain/build/binutils-2.18.50/gas -I. -I.
> -I/opt/or1200-toolchain/build/binutils-2.18.50/gas -I../bfd
> -I/opt/or1200-toolchain/build/binutils-2.18.50/gas/config
> -I/opt/or1200-toolchain/build/binutils-2.18.50/gas/../include
> -I/opt/or1200-toolchain/build/binutils-2.18.50/gas/..
> -I/opt/or1200-toolchain/build/binutils-2.18.50/gas/../bfd
>
>

-DLOCALEDIR="\"/opt/or1200-toolchain/build/tools/or32-uclinux/share/locale\""

>  -W -Wall -Wstrict-prototypes -Wmissing-prototypes -Werror -g -O2
> -c
> /opt/or1200-toolchain/build/binutils-2.18.50/gas/listing.c
> cc1: warnings being treated as errors
> /opt/or1200-toolchain/build/binutils-2.18.50/gas/listing.c: In
> function ‘print_options’:
> /opt/or1200-toolchain/build/binutils-2.18.50/gas/listing.c:1103:
> error: format not a string literal and no format arguments
> make[4]: *** [listing.o] Error 1
> The line concerned in listing.c is:
> fprintf (list_file, field_name);
> So I presume here we put in a format string like before, again wary
> that I have not much of an idea of what this is meant to do:
> fprintf (list_file, "%s", field_name);
> And now binutils-2.18.50 compiles and installs.
> GCC-4.2.2 seems to compile OK too, however the next issues with
> this
> toolchain build are to do with the linux kernel portion, but I'm
> yet
> to track down the specifics of this. I'll post more when I know.
> Anyway, i'm not sure if perhaps there's a configure option I can
> pass
> (I've seen the -Wformat-nonliteral, but I guess this makes gcc
> thrown
> an error at any of these warnings.... which is now standard
> behavior
> in gcc-4.3.2 I think) or perhaps these fixes should be in the next
> patch (or perhaps already are fixed in binutils-2.19). Anyone know
> what could be done to make this work first time, every time?
> -jb
>
>

_______________________________________________
http://www.opencores.org/mailman/listinfo/openrisc

[rich_daddio]

Hi JB,
Yes if you peruse the list you will find others have posted about
this. Yan Morvan has posted a patch to fix the mof_ofsoc build script
and we will be incorporating this soon. Also thanks for the tip on
2.19. I am hoping to upgrade our toolchain to newer versions soon as well.

Thanks,

Rich d

----- Original Message -----
From: jb at orsoc.se<jb@o...>
To:
Date: Thu Mar  5 12:16:39 CET 2009
Subject: [openrisc] Building binutils 2.18.50 on Ubuntu 8.10 (gcc
4.3.2)

> Well, I've got the answer to my own problem here.
>
> It appears that this change in behavior of gcc-4.3.2 has been
> addressed in binutils-2.19, those two lines have been changed to
> be,
> in the case of binutils-2.19/binutils/dwarf.c:
> fputs (buff + (byte_size == 4 ? 8 : 0), stdout);
> and in binutils-2.19/gas/listing.c:
> fputs (field_name, list_file);
> fputs() saves the day from nasty string literal errors.
> ----- Original Message -----
> From: jb at orsoc.se<jb at o...>
> To:
> Date: Thu Mar 5 11:25:19 CET 2009
> Subject: [openrisc] Building binutils 2.18.50 on Ubuntu 8.10 (gcc
> 4.3.2)
> > Hi all,
> >
> > I've been testing building the toolchain on the most recent
> major
> > Ubuntu release, 8.10, and it appears that binutils breaks due
> to
> > stricter enforcement of string literals being passed to things
> like
> > printf and fprintf. For instance, out of the box, binutils
> 2.18.50
> > with the appropriate patch errors out the first of two times
> on
> > binutils-2.18.50/binutils/dwarf.c like so:
> > gcc -DHAVE_CONFIG_H -I.
> > -I/opt/or1200-toolchain/build/binutils-2.18.50/binutils -I.
> -I.
> > -I/opt/or1200-toolchain/build/binutils-2.18.50/binutils
> -I../bfd
> > -I/opt/or1200-toolchain/build/binutils-2.18.50/binutils/../bfd
> >
> -I/opt/or1200-toolchain/build/binutils-2.18.50/binutils/../include
> >
> >
>
>

-DLOCALEDIR="\"/opt/or1200-toolchain/build/tools/or32-uclinux/share/locale\""

> > -Dbin_dummy_emulation=bin_vanilla_emulation -W -Wall
> > -Wstrict-prototypes -Wmissing-prototypes -Werror -g -O2 -c
> > /opt/or1200-toolchain/build/binutils-2.18.50/binutils/dwarf.c
> > cc1: warnings being treated as errors
> > /opt/or1200-toolchain/build/binutils-2.18.50/binutils/dwarf.c:
> In
> > function ‘print_dwarf_vma’:
> >
> /opt/or1200-toolchain/build/binutils-2.18.50/binutils/dwarf.c:189:
> > error: format not a string literal and no format arguments
> > make[4]: *** [dwarf.o] Error 1
> > It was configured from a recent version of the MOF script,
> like so:
> > $BUILD_TOP/$BINUTILS_VER/configure --target=or32-uclinux
> > --prefix=$BUILD_TOP/tools/or32-uclinux --disable-checking
> > --disable-Werror
> > So, the line it refers to in dwarf.c that it doesn't like is
> like
> > this:
> > printf (buff + (byte_size == 4 ? 8 : 0));
> > Now, I'm by no means one hundred percent that this is correct,
> but
> > I
> > presume there's a pre-made string it wants to pass to printf,
> so if
> > I
> > just put in a "%s" in there it will fix this error?
> Wary
> > that I may
> > have changed the code so that it doesn't do what was intended,
> I
> > can
> > at least make the file compile by changing the line at 189 to:
> > printf ("%s", buff + (byte_size == 4 ? 8 : 0));
> > The next issue when compiling is for the same reason, in the
> file
> > binutils-2.18.50/gas/listing.c:
> > gcc -DHAVE_CONFIG_H -I.
> > -I/opt/or1200-toolchain/build/binutils-2.18.50/gas -I. -I.
> > -I/opt/or1200-toolchain/build/binutils-2.18.50/gas -I../bfd
> > -I/opt/or1200-toolchain/build/binutils-2.18.50/gas/config
> > -I/opt/or1200-toolchain/build/binutils-2.18.50/gas/../include
> > -I/opt/or1200-toolchain/build/binutils-2.18.50/gas/..
> > -I/opt/or1200-toolchain/build/binutils-2.18.50/gas/../bfd
> >
> >
>
>

-DLOCALEDIR="\"/opt/or1200-toolchain/build/tools/or32-uclinux/share/locale\""

> > -W -Wall -Wstrict-prototypes -Wmissing-prototypes -Werror -g
> -O2
> > -c
> > /opt/or1200-toolchain/build/binutils-2.18.50/gas/listing.c
> > cc1: warnings being treated as errors
> > /opt/or1200-toolchain/build/binutils-2.18.50/gas/listing.c: In
> > function ‘print_options’:
> >
> /opt/or1200-toolchain/build/binutils-2.18.50/gas/listing.c:1103:
> > error: format not a string literal and no format arguments
> > make[4]: *** [listing.o] Error 1
> > The line concerned in listing.c is:
> > fprintf (list_file, field_name);
> > So I presume here we put in a format string like before, again
> wary
> > that I have not much of an idea of what this is meant to do:
> > fprintf (list_file, "%s", field_name);
> > And now binutils-2.18.50 compiles and installs.
> > GCC-4.2.2 seems to compile OK too, however the next issues
> with
> > this
> > toolchain build are to do with the linux kernel portion, but
> I'm
> > yet
> > to track down the specifics of this. I'll post more when I
> know.
> > Anyway, i'm not sure if perhaps there's a configure option I
> can
> > pass
> > (I've seen the -Wformat-nonliteral, but I guess this makes gcc
> > thrown
> > an error at any of these warnings.... which is now standard
> > behavior
> > in gcc-4.3.2 I think) or perhaps these fixes should be in the
> next
> > patch (or perhaps already are fixed in binutils-2.19). Anyone
> know > what could be done to make this work first time, every
> time? > -jb > >
>

_______________________________________________
http://www.opencores.org/mailman/listinfo/openrisc