Re: RFC: invoke ACX_LARGEFILE for gdbserver

[Pedro Alves <palves at redhat dot com>]

On 02/21/2013 07:09 PM, Tom Tromey wrote:
>>>>>> "Jan" == Jan Kratochvil <jan.kratochvil@redhat.com> writes:
> 
> Tom> An alternative approach would be to directly invoke AC_SYS_LARGEFILE.
> 
> Jan> I am not sure how it is with gdbserver and procfs on Solaris but as it does
> Jan> not seem to be an issue I would be for the more simple AC_SYS_LARGEFILE.
> 
> Here you go.

I'm OK with either approach.  Either patch is OK.

FYI, a gdbserver Solaris port was posted once:

 http://sourceware.org/ml/gdb-patches/2010-10/msg00179.html

and if we do end up splitting the gdb/gdbserver backends to a
library to be shared with gdb, we will end up needing to revisit
this at some (possibly somewhat distant) point.

I find the check for plugins support in largefile.m4 mystifying.

case "${host}" in
changequote(,)dnl
  sparc-*-solaris*|i[3-7]86-*-solaris*)
changequote([,])dnl
    # On native 32bit sparc and ia32 solaris, large-file and procfs support
    # are mutually exclusive; and without procfs support, the bfd/ elf module
    # cannot provide certain routines such as elfcore_write_prpsinfo
    # or elfcore_write_prstatus.  So unless the user explicitly requested
    # large-file support through the --enable-largefile switch, disable
    # large-file support in favor of procfs support.
    test "${target}" = "${host}" -a "x$plugins" = xno \
      && : ${enable_largefile="no"}
    ;;
esac

It'd be good to have that comment expanded, as it currently doesn't
mention the rationale for the plugins=no check at all.  I already spent
more than I'd like looking for the rationale for that, but the closest I
could find was

  http://sourceware.org/ml/binutils/2009-05/msg00402.html

which doesn't really explain it.  Doesn't it look like a build with
"--enable-plugins" on 32-bit Solaris will always be broken?

I'd guess that a --enable-targets=all build, which enables plugins
too, IIRC, may also be causing trouble on ia32 Solaris, as that's not
checked here.  Blah.  The elfcore_write_prpsinfo/elfcore_write_prstatus
issue should really be fixed by making the core generation host independent,
thus independent of procfs, but that'd still leave gdb with the same issue.
Double blah.

I saw nowhere, in either the code or in the archives, why plugins require
largefile support.
Triple blah.


-- 
Pedro Alves