Re: gdb-patches Digest 15 Feb 2001 06:20:08 -0000 Issue 549

[Jeffrey A Law <law at redhat dot com>]

  In message <982218008.28188.ezmlm@sources.redhat.com>you write:
  > From: Michael Elizabeth Chastain <chastain@cygnus.com>
  > Date: Wed, 14 Feb 2001 21:04:55 -0800
  > Message-Id: <200102150504.VAA15952@bosch.cygnus.com>
  > To: gdb-patches@sources.redhat.com
  > Subject: [PATCH] som_solib_add_solib_objfile
  > 
  > This is an hpux-specific patch.
  > 
  > About a year ago, gdb moved from fixed section names (text, data, bss)
  > to arbitary section names.  At that time, syms_from_objfile changed
  > its interface from "one address to relocate every section" to "table of
  > specific section offsets".  The hpux shared library reader was changed
  > to use the new interface, but did so in a broken way.
  > 
  > This patch changes som_solib_add_solib_objfile to do section relocation
  > properly.  This fixes the ability of gdb to call shared library functions
  > in the inferior, including the implicit "malloc" calls that gdb uses
  > when calling any function with string parameters.
  > 
  > Testing: I've tested this patch extensively with hpux 10.20 on a different
  > branch of gdb.  I've never tested it on hpux 11.00.  I haven't tested
  > it with FSF CVS gdb; the last time I built FSF CVS gdb, it could not
  > run inferior programs at all.
  > 
  > Because of the lack of testing, I'm submitting this patch to the hpux
  > maintainer, rather than asking for approval to commit it myself (but
  > I'll do that if that's what he wants).
  > 
  > Michael
  > 
  > ===
  > 
  > 2001-02-12  Michael Chastain  <chastain@redhat.com>
  > 
  > 	* gdb/somsolib.c (som_solib_add_solib_objfile): Do not use
  > 	section relocation feature of syms_from_objfile.  Do my own
  > 	section relocation, offsetting each section of the som by
  > 	either text_addr - text_link_addr or data_start.
This is fine.  Please install this patch.  Thanks for testing it on
hpux10.20, doing so saved me quite a bit of time (my 10.20 box is fairly
slow).

Also thanks for tackling the data relocation problem -- it's been bugging
me for a while.

jeff

ps.  As Michael mentioned, hpux11 gdb has gone totally schizo -- it's
hanging in the ttrace code waiting for the inferior to say "hey, I forked".
I have no idea how the ttrace code is supposed to work.