Re: 2 weeks to GDB 7.10 tentative branching date!

[Doug Evans <dje at google dot com>]

On Mon, Jun 1, 2015 at 2:37 PM, Joel Brobecker <brobecker@adacore.com> wrote:
> Hello,
>
> Just a quick reminder that we've set the tentative branching
> date for the GDB 7.10 release to June 15th, which is only
> a couple of weeks away.
>
> As in previous releases, we have a wiki page where we track all
> blocking issues before we can cut the branch:
>
>     https://sourceware.org/gdb/wiki/GDB_7.10_Release
>
> If you know of anything that we should hold the branch for, please
> let us know. There are typically regressions, important features
> that need a little extra time to be completely integrated in master,
> etc.
>
> Thank you!
> --
> Joel

It would be great to resolve pr 16253 for 7.10.
We reverted a fix in the lead up to 7.8, hoping to revisit it
afterwards, but we never did. :-(
[details in the pr]

https://sourceware.org/bugzilla/show_bug.cgi?id=16253

Maybe a two step solution would be ok?

For 7.10:

1) re-add keiths patch, with the following modifications:
   a) add a STRUCT vs VAR_DOMAIN indicator to .gdb_index
   b) bump .gdb_index version to 9
   c) modify keith's patch so that it only is enabled when either
       .gdb_index is not in use or its version >= 9
       (IOW we'd still have to keep symbol_matches_domain,
       in some form, but use it conditionally)

For later:

2) when all producers we care about can produce v9 indices,
    remove hack 1c

I'm aware of the u/i consequences of this approach
(whether examples like in 16253 would work would be
dependent upon whether the index is in use and its version).

Another thought I had is that symbol lookup has changed a
bit since 7.8. It may be possible to just reapply keith's
patch and avoid the massive perf regression differently.

Comments?