[PATCH v2] Improve building with LTO

[Dmitry V. Levin]

Hi Mark,

On Thu, Nov 04, 2021 at 01:12:29PM +0100, Mark Wielaard wrote:
> Hi Dmitry,
> 
> On Thu, 2021-11-04 at 14:23 +0300, Dmitry V. Levin wrote:
> > On Sat, Aug 28, 2021 at 12:31:43PM +0300, Dmitry V. Levin wrote:
> > > On Thu, Feb 18, 2021 at 03:38:56AM +0100, Alexander Miller via
> > > Elfutils-devel wrote:
> > > > From: Alexander Miller <alex.miller@gmx.de>
> > > > 
> > > > Use symver attribute for symbol versioning instead of .symver
> > > > assembler directive when available. Convert to use double @
> > > > syntax
> > > > for default version in all cases (required when using the
> > > > attribute).
> > > > 
> > > > Add the attributes externally_visible, no_reorder if available
> > > > when
> > > > using assembler directives to improve the situation for < gcc-10.
> > > > This is not 100% reliable, though; -flto-partition=none may still
> > > > be
> > > > needed in some cases.
> > > > 
> > > > Note that -Wno-error=stack-usage= is still needed to build with
> > > > LTO.
> > > > 
> > > > Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=24498
> > > > Signed-off-by: Alexander Miller <alex.miller@gmx.de>
> > > > ---
> > > >  lib/ChangeLog                  | 13 +++++++
> > > >  lib/eu-config.h                | 65 +++++++++++++++++++++++++++-
> > > > ------
> > > >  libdw/ChangeLog                | 11 ++++++
> > > >  libdw/dwarf_aggregate_size.c   |  4 +--
> > > >  libdw/dwarf_arrayorder.c       |  2 +-
> > > >  libdw/dwarf_bitoffset.c        |  2 +-
> > > >  libdw/dwarf_bitsize.c          |  2 +-
> > > >  libdw/dwarf_bytesize.c         |  2 +-
> > > >  libdw/dwarf_decl_column.c      |  2 +-
> > > >  libdw/dwarf_decl_file.c        |  2 +-
> > > >  libdw/dwarf_decl_line.c        |  2 +-
> > > >  libdw/dwarf_srclang.c          |  4 +--
> > > >  libdwelf/ChangeLog             |  5 +++
> > > >  libdwelf/dwelf_elf_begin.c     |  2 +-
> > > >  libdwfl/ChangeLog              |  7 ++++
> > > >  libdwfl/core-file.c            |  4 +--
> > > >  libdwfl/dwfl_module_build_id.c |  4 +--
> > > >  libdwfl/dwfl_report_elf.c      |  4 +--
> > > >  18 files changed, 107 insertions(+), 30 deletions(-)
> > > 
> > > ping?
> > 
> > FWiW, I applied this patch in ALT's elfutils package about 2 months
> > ago.
> > Thanks to -Wstack-usage fixes merged, it no longer requires
> > -Wno-error=stack-usage= to build with LTO.
> 
> Thanks. This patch was indeed one reason I kept postponing the release,
> because I didn't have have time to properly review it.
> 
> Which gcc versions have you tried this against (with/without -flto?)

I tested with gcc10 and gcc11.
I could try older versions, although I didn't feel that necessary.

> I admit I am still a bit nervous about the switch away from @@@ to just
> @ and @@. I was secretly hoping gcc would add @@@ support to the symver
> attribute. But that doesn't seem to be happening, and even if it did,
> it would be gcc 12+ only. So maybe we can include it for this release
> and just tell people they may keep the pieces if they use -flto. But it

https://sourceware.org/bugzilla/show_bug.cgi?id=27367 will likely strike
those who would build elfutils with -flto using gcc11+.

> does also impact symbol versioning for non-lto builds, so I am still a
> little hesitant. I'll try to do some tests to make sure things look ok
> with different gcc versions.

What do you mean by "it does also impact symbol versioning for non-lto
builds"?  The code for non-lto builds changes, but the versioning
should remain the same, shouldn't it?


-- 
ldv