Re: Make pretty printers version namespace dependant

[Jonathan Wakely <jwakely at redhat dot com>]

On 27/04/17 21:27 +0200, François Dumont wrote:
> On 27/04/2017 20:32, Jonathan Wakely wrote:
> > On 27/04/17 20:12 +0200, François Dumont wrote:
> > > Hi
> > >
> > >   Here is the patch to registers Printers depending on activation 
> > > of versioned namespace.
> > >
> > > 2017-04-27  François Dumont  <fdumont@gcc.gnu.org>
> > >
> > >   * python/Makefile.am [ENABLE_SYMVERS_GNU_NAMESPACE]
> > >   (user_versioned_namespace): New.
> > >   (gdb.py): Adapt target.
> > >   * python/Makefile.in: Regenerate.
> > >   * python/hook.in: Likewise.
> > >   * python/libstdcxx/v6/printers.py (add_one_template_type_printer):
> > >   Register only 1 Printer type depending on _versioned_namespace value.
> > >   (add_one_type_printer): Likewise.
> > >   (register_libstdcxx_printers): Add parameter use_versioned_namespace.
> > >   Adapt _versioned_namespace value based on it.
> > >   * testsuite/lib/gdb-test.exp (get_use_versioned_namespace): New.
> > >   (note-test): Detect if version namespace is active and pass the
> > >   information to register_libstdcxx_printers.
> > >   * testsuite/libstdc++-prettyprinters/48362.cc: Replace a regexp-test
> > >   by a note-test.
> > >
> > > Tested under Linux x86_64, with and without version namespace.
> > >
> > > Ok to commit to trunk ?
> >
> > What's the advantage?
> >
> > Again, the point of my patch was to optimise if needed. Is it needed?
>
> Oh, I thought it was obviously better to not register useless 
> Printers. But if you need a bug report or an explicit test showing 
> that it is better then I don't have it. Maybe a gdb guy could tell us 
> if it worth it.

I think I've asked t he GDB guys before about the overhead of
printers, but I'll check with them. I don't think having extra
printers registered is a problem (GDB is usually an interactive
program, so the time it takes to find a printer is not significant on
"human scales").

> Do I just keep the small cleanup on the useless usage of regexp-test 
> in 48362.cc ?

That part is fine now, thanks.


> > > Are we going to wait for gcc 7.1 before bumping version namespace ?
> >
> > Shouldn't we wait until we change something that would require bumping
> > it? :-)
>
> Ok, I'll add it to my first patch needing it.
>
> Is it the right moment to do so ?

I think changes that only affect the versioned namespace are OK for
trunk now.