Bug#778163: vim: ftbfs with GCC-5

[James McCoy]

On Mon, Mar 02, 2015 at 11:38:40PM -0500, James McCoy wrote:
> On Fri, Feb 27, 2015 at 08:34:59PM +0100, Matthias Klose wrote:
> > On 02/27/2015 08:33 PM, James McCoy wrote:
> > > On Feb 27, 2015 1:49 PM, "Matthias Klose" <doko at debian.org> wrote:
> > >>
> > >> On 02/27/2015 07:45 PM, James McCoy wrote:
> > >>> I narrowed it down to src/regexp.c.
> > >>
> > >> cool, you could use the optimize attribute, applied to a single function
> > > to
> > >> search for the miscompiled function ...
> > > 
> > > Cool. It's the peekchr function.
> > 
> > any chance to reduce this to a test case?
> 
> You mean a test case as described in
> <https://gcc.gnu.org/bugs/minimize.html>?  The specific function this is
> happening in is an internal helper function relying on global state of
> Vim's regex engine.  I don't have a minimal runtime test case, let alone
> a minimal piece of code to reproduce the problem.

Looks like someone else may have found time to do this.  I was checking
the GCC 5 regression list today and found this tree-opt bug[0] related
to miscompiling Vim.

[0]: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65418

Their example code is from the peekchr function.  Based on that, I'm
reassigning this to gcc.

Cheers,
-- 
James
GPG Key: 4096R/331BA3DB 2011-12-05 James McCoy <jamessan at debian.org>