parallel builds failing in bfd

parallel builds failing in bfd

[David Carlton]

Every once in a while, when I do a parallel build from the top level,
I get a failure that stems from bfd.h being built incorrectly.  Do
other people run into this problem?  I'm not at all familiar with
bfd's Makefile.in, so I don't know offhand what might be causing this
problem.

David Carlton
carlton@math.stanford.edu

Re: parallel builds failing in bfd

[Daniel Jacobowitz]

On Wed, May 28, 2003 at 03:01:48PM -0700, David Carlton wrote:
> Every once in a while, when I do a parallel build from the top level,
> I get a failure that stems from bfd.h being built incorrectly.  Do
> other people run into this problem?  I'm not at all familiar with
> bfd's Makefile.in, so I don't know offhand what might be causing this
> problem.

It's an autoconf 2.13 bug unfortunately.  It will be fixed after the
2.5x transition, which is starting to gain a little momentum now.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer

Re: parallel builds failing in bfd

[Nicholas Wourms]

Daniel Jacobowitz wrote:
> On Wed, May 28, 2003 at 03:01:48PM -0700, David Carlton wrote:
> 
>>Every once in a while, when I do a parallel build from the top level,
>>I get a failure that stems from bfd.h being built incorrectly.  Do
>>other people run into this problem?  I'm not at all familiar with
>>bfd's Makefile.in, so I don't know offhand what might be causing this
>>problem.
> 
> 
> It's an autoconf 2.13 bug unfortunately.  It will be fixed after the
> 2.5x transition, which is starting to gain a little momentum now.
> 

BTW, are there any plans to transition to Automake-1.7.X & 
Libtool-1.5.X, now that they have stabilized?  It sure would be nice to 
finally have a consolidated, streamlined build without all this leftover 
Cygnus tree building crap.  Yes, I dare say to ditch the current cruft 
which is polluting Makefile.in's.  There should be no reason why the 
entire configury can't happen in one shot.  Yes, I am aware that it must 
support mixed trees, which is why one would put conditional guards 
around subproject macros, running them only iff said subproject 
directory is present in the source tree.  It would save time and, for 
some, resources [since repeatedly calling the Autom4te/Automake perl 
functions (when checking for autoheader, etc.) tends to hog resources]. 
  Not to mention the fact that {in-tree,parallel,distcc} building is 
*much* easier to implement and less prone to break with the latest 
versions of these tools.  The biggest advantage, by far, would be to 
bring subproject Makefile templates inline with some sort of general 
standard, which is generally enforced by the current Automake/Autoconf 
macros (whereas previous versions of autoconf/automake were rather lax).

Just my $0.02...  Feel free to disagree :-)

Cheers,
Nicholas

Re: parallel builds failing in bfd

[DJ Delorie]

> BTW, are there any plans to transition to Automake-1.7.X &
> Libtool-1.5.X, now that they have stabilized?  It sure would be nice
> to finally have a consolidated, streamlined build without all this
> leftover Cygnus tree building crap.  Yes, I dare say to ditch the
> current cruft which is polluting Makefile.in's.

Are you willing to do the work?  We got the toplevel files
autoconfiscated only due to a tremendous amount of volunteer time (by
Nate), and any further work is probably also going to have to wait for
someone to volunteers to do it.

Re: parallel builds failing in bfd

[amodra]

On Wed, May 28, 2003 at 07:59:05PM -0400, DJ Delorie wrote:
> Nate), and any further work is probably also going to have to wait for
> someone to volunteers to do it.

Various people have volunteered to do the work in the past, eg. see
http://sources.redhat.com/ml/binutils/2003-01/msg00143.html
It's more a matter of having _maintainers_ volunteer the time to
review and/or be willing to risk committing a patch that will likely
have some initial faults.

-- 
Alan Modra
IBM OzLabs - Linux Technology Centre

unsubsribe

[Gitesh Kulkarni]

unsubsribe

Re: parallel builds failing in bfd

[H. J. Lu]

On Wed, May 28, 2003 at 06:48:32PM -0400, Nicholas Wourms wrote:
> Daniel Jacobowitz wrote:
> > On Wed, May 28, 2003 at 03:01:48PM -0700, David Carlton wrote:
> > 
> >>Every once in a while, when I do a parallel build from the top level,
> >>I get a failure that stems from bfd.h being built incorrectly.  Do
> >>other people run into this problem?  I'm not at all familiar with
> >>bfd's Makefile.in, so I don't know offhand what might be causing this
> >>problem.
> > 
> > 
> > It's an autoconf 2.13 bug unfortunately.  It will be fixed after the
> > 2.5x transition, which is starting to gain a little momentum now.
> > 
> 
> BTW, are there any plans to transition to Automake-1.7.X & 
> Libtool-1.5.X, now that they have stabilized?  It sure would be nice to 

FYI, the current binutils/gdb tree doesn't work with

# make install DESTDIR=foo

when shared library is enabled due to the libtool bug. I put a kludge
in my Linux binutils. I hope it gets fixed soon.


H.J.