gcc-patches@gcc.gnu.org
[Top] [All Lists]

Re: [PATCH/RFC] PR other/22313: Hot/cold sections vs. dwarf2 (take 2)

Subject: Re: [PATCH/RFC] PR other/22313: Hot/cold sections vs. dwarf2 take 2
From: Roger Sayle
Date: Mon, 18 Sep 2006 13:37:48 -0600 MDT
Hi David,

On Mon, 18 Sep 2006, David Daney wrote:
> > 2006-08-31  Roger Sayle  <roger@xxxxxxxxxxxx>
> >
> >     PR other/22313
> >     * dwarf2out.c (add_fde_cfi): Use a set_loc if the current label is
> >     NULL, otherwise use an advance_loc4 to adjust relative to the
> >     current label.
> >     (output_cfi) <DW_CFA_set_loc>: Update the current label.
> >     (dwarf2out_switch_text_section): Reset the current label to avoid
> >     using advance_loc4 over section boundaries.
> >
> >
>
> This breaks exception handling on mipsel-linux-gnu.  I did a regression
> hunt and r116604 (this patch) fails, where as r116603 works.
>
> The problem is that with this patch applied the first instruction in a
> FDE is typically DW_CFA_set_loc, before the patch it was DW_CFA_advance_loc.

Grr!  Thanks for tracking down the change that broke EH on MIPS!

I think I can provide a temporary work-around for this problem (to
resolve the immediate regression) by initializing the current label
to the start of the function.  This'll mean that we'll resume using
advance_loc as the first instruction in the FDE.  However, profiled
bootstrap and/or hot/cold partitioning will still continue to introduce
DW_CFA_set_loc, which presumably can't be handled by current binutils.

Not ideal, but it will restore libgcj whilst the necessary bintuils
functionality is added.

Sorry for the inconvenience,

Roger
--

<Prev in Thread] Current Thread [Next in Thread>