Re: binutils newer than 2.9.1

Re: binutils newer than 2.9.1

[Ian Lance Taylor]

   Date: Mon, 07 Jun 1999 06:04:18 +0000
   From: Brendan Simon <brendan@dgs.monash.edu.au>

   Where can I find binutils releases that are newer version 2.9.1 ?

There are no official binutils releases newer than 2.9.1.  The CVS
binutils sources happen to report themselves as 2.9.4 currently, but
that is strictly an internal number.  There is no single 2.9.4
release.

   I read in one of Mumit Khan's cross-compiler howtos that 2.9.4 is
   recommended for building the cygwin32/mingw32 compilers.

The cygwin releases include binutils snapshots which have been tested
for cygwin.  Probably the latest cygwin release included a binutils
snapshot marked 2.9.4 (it's been 2.9.4 for a while).  But, again,
there is no single 2.9.4 release.

   The official binutils site (sourceware.cygnus.com as far as I am aware)
   has 2.9.1as the latest release.  This has been around for quite awhile
   and I'm sure there must have been further releases.  Has the site been
   kept up to date ?

The site is more or less up to date.

I'm afraid we are very far behind in actually making an official
release.  I still have a stack of patches and bug reports I need to
get through, although I have been making a little progress recently.

Ian

Re: binutils newer than 2.9.1

[Mumit Khan]

On Tue, 8 Jun 1999, Brendan Simon wrote:

> Mumit Khan wrote:
> 
> > To get what I call binutils-2.9.4 in my howto, you can get either get
> > the whole Cygwin source tree [1] (big!), or just get the pieces that
> > form binutils separately [2] that you can combine to re-create the 2.9.4
> > tree.

[...]

> 
> I assume these files have other updates for targets other than
> cygwin32/mingw32.  Is that correct ?
> Is this site likely to be maintained/updated regularly or do you think that I
> am better off using CVS and checking out snapshots that are known to be
> reasonably reliable ?

I'm in no position to judge whether the binutils packaged with Cygwin
works well with targets other than pe-coff or not. I only use this for
the various x86-win32 targets that I use, and use binutils-2.9.1 + patches
that are on the net for other targets (including ppc-eabi).

I do remember that I had some trouble with ppc-eabi when using binutils
2.9.4 (forward labels didn't have the right offset or some such thing),
and ended up using 2.9.1.

If you're supporting many targets, getting it out of CVS is probably the
way to go (don't forget to run the testsuite!).

Regards,
Mumit

binutils newer than 2.9.1

[Brendan Simon]

Where can I find binutils releases that are newer version 2.9.1 ?
I read in one of Mumit Khan's cross-compiler howtos that 2.9.4 is
recommended for building the cygwin32/mingw32 compilers.  There is a
reference to where to obtain these but I would like to latest tested
release when building all the cross-compilers I am interested in (mainly
powerpc and m68k).  I'd prefer not to use CVS as this might be too
bleeding-edge for my purposes.

The official binutils site (sourceware.cygnus.com as far as I am aware)
has 2.9.1as the latest release.  This has been around for quite awhile
and I'm sure there must have been further releases.  Has the site been
kept up to date ?

The main sourceware.cygnus.com web page needs a link to the binutils
page in the left frame.  There is a little message in the main frame
saying that binutils is now hosted at soureware but howlong will that
message be there for ?

Brendan Simon.

Re: binutils newer than 2.9.1

[Ian Lance Taylor]

   Date: Wed, 09 Jun 1999 09:00:02 -0400
   From: "Charles Gauthier" <Charles.Gauthier@iit.nrc.ca>

   We are using binutils 2.9.4 for m68k-rtemself (really elf32-m68k).

I just want to caution people against referring to any set of sources
as binutils 2.9.4.  There is no single set of sources which that
describes.  Some binutils which identify themselves as 2.9.4 will be
very buggy and unusable.  Some may work fine.

Ian

Re: binutils newer than 2.9.1

[Brendan Simon]

Ian Lance Taylor wrote:

>    Date: Mon, 07 Jun 1999 06:04:18 +0000
>    From: Brendan Simon <brendan@dgs.monash.edu.au>
>
>    Where can I find binutils releases that are newer version 2.9.1 ?
>
> There are no official binutils releases newer than 2.9.1.  The CVS
> binutils sources happen to report themselves as 2.9.4 currently, but
> that is strictly an internal number.  There is no single 2.9.4
> release.
>
>    I read in one of Mumit Khan's cross-compiler howtos that 2.9.4 is
>    recommended for building the cygwin32/mingw32 compilers.
>
> The cygwin releases include binutils snapshots which have been tested
> for cygwin.  Probably the latest cygwin release included a binutils
> snapshot marked 2.9.4 (it's been 2.9.4 for a while).  But, again,
> there is no single 2.9.4 release.
>
>    The official binutils site (sourceware.cygnus.com as far as I am aware)
>    has 2.9.1as the latest release.  This has been around for quite awhile
>    and I'm sure there must have been further releases.  Has the site been
>    kept up to date ?
>
> The site is more or less up to date.
>
> I'm afraid we are very far behind in actually making an official
> release.  I still have a stack of patches and bug reports I need to
> get through, although I have been making a little progress recently.
>
> Ian

If I want later versions of binutils (I am unsure of how necessary this is)
then it would appear that CVS would be the best option and to checkout
snapshots that are known to be reasonably reliable.  Would you suggest this
as a wise thing or to just stick with the stock standard 2.9.1 ?  Maybe the
cygwin sources of binutils might be the best option ?  My main targets of
interest are powerpc-eabi, powerpc-linux and m68k-???

Brendan Simon.

Re: binutils newer than 2.9.1

[Brendan Simon]

Mumit Khan wrote:

> To get what I call binutils-2.9.4 in my howto, you can get either get
> the whole Cygwin source tree [1] (big!), or just get the pieces that
> form binutils separately [2] that you can combine to re-create the 2.9.4
> tree.
>
> [1] ftp://sourceware.cygnus.com/pub/cygwin/cygwin-b20/
> [2] ftp://sourceware.cygnus.com/pub/cygwin/cygwin-b20/src-by-top-dir/
>
> Here's what I had posted to the Cygwin mailing list a while back:
>
> Date:    Thu, 18 Feb 1999 14:40:35 CST
> cc:      cygwin mailing list <cygwin@sourceware.cygnus.com>
> Subject: Re: How to compile binutils
>
> Thanks to DJ's addition of toplevel.tar.bz2, it's very complete.
> Here's a list of files needed for binutils-2.9.4 (I just wrote this
> up yesterday for my mingw-cross-howto guide, so you're in luck):
>
>   bfd.tar.bz2        gas.tar.bz2        ld.tar.bz2        toplevel.tar.bz2
>   binutils.tar.bz2   gprof.tar.bz2      libiberty.tar.bz2
>   config.tar.bz2     include.tar.bz2    opcodes.tar.bz2
>   etc.tar.bz2        intl.tar.bz2       texinfo.tar.bz2

I assume these files have other updates for targets other than
cygwin32/mingw32.  Is that correct ?
Is this site likely to be maintained/updated regularly or do you think that I
am better off using CVS and checking out snapshots that are known to be
reasonably reliable ?

Thanks,
Brendan Simon.

Re: binutils newer than 2.9.1

[Mumit Khan]

Brendan Simon <brendan@dgs.monash.edu.au> writes:
> Where can I find binutils releases that are newer version 2.9.1 ?
> I read in one of Mumit Khan's cross-compiler howtos that 2.9.4 is
> recommended for building the cygwin32/mingw32 compilers.  There is a
> reference to where to obtain these but I would like to latest tested
> release when building all the cross-compilers I am interested in (mainly
> powerpc and m68k).  I'd prefer not to use CVS as this might be too
> bleeding-edge for my purposes.

(I assume you've already seen Ian's response to your questions)

To get what I call binutils-2.9.4 in my howto, you can get either get
the whole Cygwin source tree [1] (big!), or just get the pieces that
form binutils separately [2] that you can combine to re-create the 2.9.4 
tree. 

[1] ftp://sourceware.cygnus.com/pub/cygwin/cygwin-b20/
[2] ftp://sourceware.cygnus.com/pub/cygwin/cygwin-b20/src-by-top-dir/

Here's what I had posted to the Cygwin mailing list a while back:

Date:    Thu, 18 Feb 1999 14:40:35 CST
cc:      cygwin mailing list <cygwin@sourceware.cygnus.com>
Subject: Re: How to compile binutils

Thanks to DJ's addition of toplevel.tar.bz2, it's very complete.
Here's a list of files needed for binutils-2.9.4 (I just wrote this
up yesterday for my mingw-cross-howto guide, so you're in luck):

  bfd.tar.bz2        gas.tar.bz2        ld.tar.bz2        toplevel.tar.bz2
  binutils.tar.bz2   gprof.tar.bz2      libiberty.tar.bz2
  config.tar.bz2     include.tar.bz2    opcodes.tar.bz2
  etc.tar.bz2        intl.tar.bz2       texinfo.tar.bz2

After you get these, unpack each one in the same place and you'll have
everything needed to rebuild binutils.

Regards,
Mumit

Re: binutils newer than 2.9.1

[Ian Lance Taylor]

   Date: Tue, 08 Jun 1999 00:22:14 +0000
   From: Brendan Simon <brendan@dgs.monash.edu.au>

   If I want later versions of binutils (I am unsure of how necessary this is)
   then it would appear that CVS would be the best option and to checkout
   snapshots that are known to be reasonably reliable.  Would you suggest this
   as a wise thing or to just stick with the stock standard 2.9.1 ?  Maybe the
   cygwin sources of binutils might be the best option ?  My main targets of
   interest are powerpc-eabi, powerpc-linux and m68k-???

I don't have a good answer to this question.  binutils 2.9.1 is
certainly better tested than the CVS snapshots.  For GNU/Linux, H.J.'s
snapshot releases are well tested.  For anything else, I guess you
just have to balance newer sources with lack of testing.  I don't know
how to resolve that balance for you.

If you have specific bugs that you want fixed, consider looking
through the CVS sources and picking out the bug fixes.  If you can't
do that, using the CVS sources may be too risky for you, since you may
run into bugs you can't fix.

Ian

Re: binutils newer than 2.9.1

[Charles Gauthier]

We are using binutils 2.9.4 for m68k-rtemself (really elf32-m68k). We have not yet
identified any problems with this target. The code was checked out of the CVS
repository two weeks ago.

The only difference I saw between 2.9.1 and .2.9.4 as an end-user is in the linker
scripts. 2.9.4 has new directives which were used extensively in the sample elf
linker scripts. Specifically, SORT, KEEP, and CONSTRUCTORS were introduced. For
elf targets, we elected to stick with 2.9.4 and to use the new linking directives.

I have no experience with 2.9.4 and powerpc, other than building the toolsuite for
powerpc-rtems from the cygwin source. It built! Once my colleague finishes the
port of RTEMS to the MBX860, I intend to try 2.9.4 from the CVS repository, just
to keep all my cross development tools at the same version level. I believe she is
doing all of her work with 2.9.1.

I think that powerpc-rtems, powerpc-eabi and powerpc-linux are all the same in the
sense that they produce elf32-powerpc code. The difference is only in the default
compilation flags and default libraries and startup files. Any bug in binutils
will likely show up for all three targets.

I have no strong opinion about using 2.9.1 or 2.9.4. Neither one has caused us
problems (that we know about).

Ian Lance Taylor wrote:

>    Date: Tue, 08 Jun 1999 00:22:14 +0000
>    From: Brendan Simon <brendan@dgs.monash.edu.au>
>
>    If I want later versions of binutils (I am unsure of how necessary this is)
>    then it would appear that CVS would be the best option and to checkout
>    snapshots that are known to be reasonably reliable.  Would you suggest this
>    as a wise thing or to just stick with the stock standard 2.9.1 ?  Maybe the
>    cygwin sources of binutils might be the best option ?  My main targets of
>    interest are powerpc-eabi, powerpc-linux and m68k-???
>
> I don't have a good answer to this question.  binutils 2.9.1 is
> certainly better tested than the CVS snapshots.  For GNU/Linux, H.J.'s
> snapshot releases are well tested.  For anything else, I guess you
> just have to balance newer sources with lack of testing.  I don't know
> how to resolve that balance for you.
>
> If you have specific bugs that you want fixed, consider looking
> through the CVS sources and picking out the bug fixes.  If you can't
> do that, using the CVS sources may be too risky for you, since you may
> run into bugs you can't fix.
>
> Ian

--
Charles-Antoine Gauthier
Research Officer
Software Engineering Group
Institute for Information Technology
National Research Council of Canada
Ottawa, ON, Canada
K1A 0R6