Re: [patch] Fix for PIE with both -e and --core in use

[Mark Wielaard <mjw at redhat dot com>]

On Wed, 2012-11-28 at 17:31 +0100, Jan Kratochvil wrote:
> Application should be able to write its own parser
> and this is just one its instance (it could not anyway but that is a different
> patch "[patch 3/4] unwinder: Provide dwfl_core_filename_report").

Yeah. You are a few steps ahead of me.

> But in practice if you used dwfl_standard_argp then an application had to give
> up using USERDATA and there was no other field for the application left, as
> application cannot modify dwfl_standard_argp to share the USERDATA field some
> way, so an application would have to copy+modify / reimplement
> dwfl_standard_argp.  Which is not great just to save a Dwfl_Module field.

So in the end it is just an dwfl_standard_argp () implementation detail.
OK, I cannot think of a better way than using some internal libwfl
field. And if we do think of something better we now have testcases.

Still one suggestion. There can be only one executable for a core based
Dwfl, can't there? So if we are using libdwfl internals anyway it seems
better to store the state in the Dwfl struct instead of reserving a
field in each Dwfl_Module struct.

And one nit. The libdwfl/ChangeLog has a line that is way too long.

Thanks,

Mark