Re: [PATCH] gdb: reset/recompute objfile section offsets in reread_symbols

[Simon Marchi <simon.marchi@efficios.com>]

On 2020-05-20 3:36 p.m., Tom Tromey wrote:
>>>>>> "Simon" == Simon Marchi via Gdb-patches <gdb-patches@sourceware.org> writes:
> 
> Simon> This patch fixes this by making reread_symbols clear the
> Simon> `section_offsets` vector, and re-create it by calling `sym_offsets`.
> Simon> This is analogous to what syms_from_objfile_1 does.  I didn't seem
> Simon> absolutely necessary, but I also made it clear the various
> Simon> `sect_index_*` fields, since their values no longer make sense (they
> Simon> describe the old executable, and are indices in the now cleared
> Simon> sections/section_offsets arrays).
> 
> This makes sense to me.

Thanks, I'll push the patch then.

> Simon> I don't really like the approach taken by reread_symbols, trying to
> Simon> reset everything manually on the objfile object, instead of, for
> Simon> example, creating a new one from scratch.
> 
> A long time ago, Jan had a patch to do exactly this.
> I couldn't easily find it; there was some opposition to it at the time,
> but I still think it's a better approach.

Here's the thread:

https://sourceware.org/legacy-ml/gdb-patches/2009-06/msg00679.html

Simon