Re: [PATCH] Scan for Mach-O start address in LC_MAIN and properly check bfd_mach_o_scan_start_address's return value

[David Albert <davidbalbert at gmail dot com>]

On Tue, Nov 27, 2012 at 5:24 AM, Joel Brobecker <brobecker@adacore.com> wrote:
>
> > > I recommend you do the paperwork now, because it does take a few
> > > weeks to complete.
> >
> > Ok, I'm happy to start this now. Someone (Nick?) let me know what I
> > need to do.
>
> For the record, I have sent David the form to get started.

Thanks!

>
>
> > > You'll need to characterize what the user-visible problems are
> > > that your patch is solving. If deemed critical and assessed as
> > > completely safe, we can hold the release for a few more days.
> >
> > It would be nice to see Tristan's changes to BFD from November 14th
> > make it into 7.5.1. Without them you get a whole bunch of errors every
> > time you run GDB (https://gist.github.com/4137073). You can continue
> > working after that, but it's ugly and gives the impression that GDB is
> > somehow broken. There are also some functional changes regarding how
> > these new load commands are handled, but I don't know how they affect
> > GDB.
>
> I just talked it over with Tristan, and we think that it's better to
> keep these changes out of the gdb-7.5 branch. We'll claim Mountain
> Lion support with the GDB 7.6 release, which is only 2-3 months away.

That's fine. To get good OS X support in GDB 7.6, it would also be
nice to fix http://sourceware.org/bugzilla/show_bug.cgi?id=14870. The
bug is not OS X specific, but clang is the default compiler on 10.7
and 10.8, so it's probably more important for darwin than for other
platforms. I'm happy to help with this however I can, but it's pretty
far from the stuff I know well.

>
>
> > I don't know enough about GDB yet to know what user-visible problems
> > my patch fixes. Without it, abfd->start_address can be NULL when it
> > shouldn't be for binaries compiled on OS X 10.8 (possibly 10.7 too,
> > but I don't have an install to check).
>
> What I am wondering is what problem triggered you to look at the code.
> But it looks like the answer to that question isn't urgent anymore,
> as Tristan isn't going to be able to re-write a new version of your
> patch in time for the 7.5.1 release.

I was originally working on a patch to fix the same issue
(unrecognized load commands) that Tristan ended up fixing. When I saw
his patch it trunk, I compared it with what I had been working on and
submitted a patch with the differences.

Best,
-David

>
>
> --
> Joel