This is the mail archive of the
gdb@sourceware.cygnus.com
mailing list for the GDB project.
Re: find_minsym_and_objfile, skip_hurd_resolver...
- To: scottb at netwinder dot org
- Subject: Re: find_minsym_and_objfile, skip_hurd_resolver...
- From: Mark Kettenis <kettenis at wins dot uva dot nl>
- Date: Thu, 11 May 2000 18:50:31 +0200 (MET DST)
- CC: gdb at sourceware dot cygnus dot com
- References: <391ADA41.4E6925A2@netwinder.org>
Date: Thu, 11 May 2000 12:05:21 -0400
From: Scott Bambrough <scottb@netwinder.org>
I recently made some fixes to get gdb on ARM Linux to work
correctly with shared libraries. One of the things I actually did
was copy find_minsym_and_objfile(), skip_hurd_resolver() and
i386_linux_skip_solib_resolver() into arm-linux-tdep.c. Other than
renaming i386_linux_skip_solib_resolver() I made no changes to the
code.
I'm a wee bit surprised that you actually need that code. I mean, I'm
not using it for the Hurd (which uses the same dynamic linker, where
did you think the name "skip_hurd_resolver" came from :-)), and yet
the Hurd seems to be fine without it. Not that I think the code
should be scrapped, since I believe it provides a useful optimization
(it prevents some single stepping through the resolver code I
believe), and I'm not entirely sure that it's really not needed.
I don't really like this duplication of code; I'd prefer to share
the code between the architechures. At present I don't see a
simple way of doing that other than creating a new file for the
three functions. Right now I'm inclined to check in what I have in
the 5.0 branch and wait and seek a neater solution in the mainline.
Does anyone have any other suggestions?
I already put a comment in the code saying that I suspected that the
could should really be in a target-dependent file rather than a native
file. And I suppose I could use it for the Hurd too, so putting it in
a seperate file would be a good idea.
Mark