Re: [PATCH 0/4] bitpos expansion summary reloaded

[Jan Kratochvil <jan.kratochvil@redhat.com>]

On Wed, 24 Oct 2012 20:32:52 +0200, Tom Tromey wrote:
> >>>>> "Jan" == Jan Kratochvil <jan.kratochvil@redhat.com> writes:
> 
> Jan> (1) Check in the patchset as is while it is known not all type safety
> Jan>     regressions have been caught.
> 
> Jan> (2) Fix all -Wconversion warnings, either by cast or by type extension,
> Jan>     depending on the case.  But this can be done anytime later.
> 
> I think we should start with (1).
> 
> My rationale is that I consider the current patch set an improvement.
> It may not be perfect, but it doesn't hurt anything, and I think the
> various threads have shown that perfecting it as a precondition for
> checking it in is too much to ask.

To restate the perfection precondition reason - it was required before it has
been found out for the really right solution we need to do full (2) anyway.
Trying to automatically find all TYPE_LENGTH-derived values truncations would
be impossible anymore after checking-in partial (1) and without the (2) fix.


> This patch may make gdb uglier, in that we'll most likely add many new
> casts to the code.  I think the resulting additional safety is probably
> worth it.

Yes.


> I'm curious what your opinion is.  I didn't see it in your message.

I think (2) - fixing -Wconversion should be done for GDB.  With GDB there only
always remains a question what is a priority of this task.

This whole work has one flaw that GDB cannot access pieces of inferior objects.
Therefore most of naive user operations with >4GB inferior objects will fail
anyway.  Which reminds priorities of GDB TODO items.


Thanks,
Jan