[Bug libstdc++/83077] sso-string @ gnu-versioned-namespace.

[Bug libstdc++/83077] sso-string @ gnu-versioned-namespace.

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83077

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|9.3                         |9.4

--- Comment #4 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
GCC 9.3.0 has been released, adjusting target milestone.

[Bug libstdc++/83077] sso-string @ gnu-versioned-namespace.

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83077

Jonathan Wakely <redi at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|9.4                         |11.0

[Bug libstdc++/83077] sso-string @ gnu-versioned-namespace.

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83077

Jonathan Wakely <redi at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|11.0                        |---

[Bug libstdc++/83077] sso-string @ gnu-versioned-namespace.

[fdumont at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83077

François Dumont <fdumont at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Assignee|unassigned at gcc dot gnu.org      |fdumont at gcc dot gnu.org
             Status|NEW                         |ASSIGNED

[Bug libstdc++/83077] sso-string @ gnu-versioned-namespace.

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83077

Jonathan Wakely <redi at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |patch
   Last reconfirmed|2017-11-20 00:00:00         |2023-08-10

--- Comment #5 from Jonathan Wakely <redi at gcc dot gnu.org> ---
New patch dropped:
https://gcc.gnu.org/pipermail/gcc-patches/2023-August/626925.html

[Bug libstdc++/83077] sso-string @ gnu-versioned-namespace.

[iains at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83077

--- Comment #6 from Iain Sandoe <iains at gcc dot gnu.org> ---
is there a version available for testing, rebased to trunk (the one I saw from
Aug 19th pretty much fails to apply for most entries)?

[Bug libstdc++/83077] sso-string @ gnu-versioned-namespace.

[fdumont at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83077

--- Comment #7 from François Dumont <fdumont at gcc dot gnu.org> ---
Sure, if you follow the email thread you'll see my latest patch:

https://gcc.gnu.org/pipermail/gcc-patches/2023-August/628399.html

[Bug libstdc++/83077] sso-string @ gnu-versioned-namespace.

[iains at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83077

--- Comment #8 from Iain Sandoe <iains at gcc dot gnu.org> ---
(In reply to François Dumont from comment #7)
> Sure, if you follow the email thread you'll see my latest patch:
> 
> https://gcc.gnu.org/pipermail/gcc-patches/2023-August/628399.html

Well, I thought I have the right one, but unfortunately, that no longer applies
to trunk (or somehow I'm not getting the right attachment).

[Bug libstdc++/83077] sso-string @ gnu-versioned-namespace.

[iains at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83077

--- Comment #9 from Iain Sandoe <iains at gcc dot gnu.org> ---
(In reply to Iain Sandoe from comment #8)
> (In reply to François Dumont from comment #7)
> > Sure, if you follow the email thread you'll see my latest patch:
> > 
> > https://gcc.gnu.org/pipermail/gcc-patches/2023-August/628399.html
> 
> Well, I thought I have the right one, but unfortunately, that no longer
> applies to trunk (or somehow I'm not getting the right attachment).

OK. So hopefully, now I had the right version (which applied) I tried a
configuration :

--disable-libstdcxx-dual-abi --with-default-libstdcxx-abi=new

the build failed for me in stage 1 target build  with :

duplicate symbol 'typeinfo name for std::basic_ostringstream<wchar_t,
std::char_traits<wchar_t>, std::allocator<wchar_t> >' in:
    .libs/libstdc++.lax/libc++11convenience.a/cow-sstream-inst.o
    .libs/libstdc++.lax/libc++11convenience.a/sstream-inst.o
<snip>
duplicate symbol 'typeinfo for std::basic_stringbuf<char,
std::char_traits<char>, std::allocator<char> >' in:
    .libs/libstdc++.lax/libc++11convenience.a/cow-sstream-inst.o
    .libs/libstdc++.lax/libc++11convenience.a/sstream-inst.o
ld: 16 duplicate symbols for architecture x86_64

[Bug libstdc++/83077] sso-string @ gnu-versioned-namespace.

[fdumont at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83077

--- Comment #10 from François Dumont <fdumont at gcc dot gnu.org> ---
This is because you are facing the PR65762 issue. I just attached a path
proposal to it that you need to apply too to be able to run your test. You'll
be even able to simply use --disable-libstdcxx-dual-abi cause I made cxx11 abi
the default in this case.

[Bug libstdc++/83077] sso-string @ gnu-versioned-namespace.

[iains at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83077

--- Comment #11 from Iain Sandoe <iains at gcc dot gnu.org> ---
(In reply to François Dumont from comment #10)
> This is because you are facing the PR65762 issue. I just attached a path
> proposal to it that you need to apply too to be able to run your test.
> You'll be even able to simply use --disable-libstdcxx-dual-abi cause I made
> cxx11 abi the default in this case.

Yes, with the second patch applied, that works for me also.

Does gnu-versioned-namespace work for you with these two patches applied?
(I have a patch to enable versioned namespace on Darwin, which is very similar
to the GNU one, but it no longer builds)..

[Bug libstdc++/83077] sso-string @ gnu-versioned-namespace.

[iains at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83077

--- Comment #12 from Iain Sandoe <iains at gcc dot gnu.org> ---
(In reply to Iain Sandoe from comment #11)
> (In reply to François Dumont from comment #10)
> > This is because you are facing the PR65762 issue. I just attached a path
> > proposal to it that you need to apply too to be able to run your test.
> > You'll be even able to simply use --disable-libstdcxx-dual-abi cause I made
> > cxx11 abi the default in this case.
> 
> Yes, with the second patch applied, that works for me also.
> 
> Does gnu-versioned-namespace work for you with these two patches applied?
> (I have a patch to enable versioned namespace on Darwin, which is very
> similar to the GNU one, but it no longer builds)..

FAOD, by which I mean "--enable-symvers=gnu-versioned-namespace" with no other
configure options (which AFAICT disables dual ABI and enables the new one) ...
I have more-or-less copied the gnu case for Darwin, but perhaps missed some
case?

[Bug libstdc++/83077] sso-string @ gnu-versioned-namespace.

[iains at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83077

--- Comment #13 from Iain Sandoe <iains at gcc dot gnu.org> ---
(In reply to Iain Sandoe from comment #12)
> (In reply to Iain Sandoe from comment #11)
> > (In reply to François Dumont from comment #10)
> > > This is because you are facing the PR65762 issue. I just attached a path
> > > proposal to it that you need to apply too to be able to run your test.
> > > You'll be even able to simply use --disable-libstdcxx-dual-abi cause I made
> > > cxx11 abi the default in this case.
> > 
> > Yes, with the second patch applied, that works for me also.
> > 
> > Does gnu-versioned-namespace work for you with these two patches applied?
> > (I have a patch to enable versioned namespace on Darwin, which is very
> > similar to the GNU one, but it no longer builds)..

It looks like this was a merge artefact, which I resolved and now it builds - I
have some testsuite fails to examine.

[Bug libstdc++/83077] sso-string @ gnu-versioned-namespace.

[fdumont at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83077

--- Comment #14 from François Dumont <fdumont at gcc dot gnu.org> ---
Good news then.

On my side I only had some failures due to a faulty friend declaration in
gnu-versioned-namespace mode in <format> for which I've submitted a patch:
https://gcc.gnu.org/pipermail/libstdc++/2023-August/056560.html

[Bug libstdc++/83077] sso-string @ gnu-versioned-namespace.

[iains at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83077

--- Comment #15 from Iain Sandoe <iains at gcc dot gnu.org> ---
(In reply to Iain Sandoe from comment #13)
> (In reply to Iain Sandoe from comment #12)
> > (In reply to Iain Sandoe from comment #11)
> > > (In reply to François Dumont from comment #10)
> > > > This is because you are facing the PR65762 issue. I just attached a path
> > > > proposal to it that you need to apply too to be able to run your test.
> > > > You'll be even able to simply use --disable-libstdcxx-dual-abi cause I made
> > > > cxx11 abi the default in this case.

> It looks like this was a merge artefact, which I resolved and now it builds
> - I have some testsuite fails to examine.

With both patches applied (on top of trunk from yesterday) on both Linux and
Darwin I am seeing regressions in the C++ and libstdc++ test suites.  For the
darwin case, I could perhaps have another merge error - but the Linux case has
only your two patches and configured with
--enable-symvers=gnu-versioned-namespace (only).

---

many of the libstdc++ fails are of this form:

/home/iains/gcc-master/bld-patched/x86_64-pc-linux-gnu/32/libstdc++-v3/include/format:3519:
error: 'std::__format::_Arg_store<_Context, _Args>::_Arg_store(_Tp& ...) [with
_Tp = {const std::chrono::time_point<std::chrono::system_clock,
std::chrono::duration<long long int, std::ratio<60, 1> > >}; _Context =
std::basic_format_context<std::__format::_Sink_iter<char>, char>; _Args =
{std::basic_format_arg<std::basic_format_context<std::__format::_Sink_iter<char>,
char> >::handle}]' is private within this context

many of the c++ fails are of this form:

contracts-tmpl-spec1.C:(.text+0x6f): undefined reference to
`handle_contract_violation(std::experimental::contract_violation const&)'

[Bug libstdc++/83077] sso-string @ gnu-versioned-namespace.

[iains at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83077

--- Comment #16 from Iain Sandoe <iains at gcc dot gnu.org> ---
(In reply to François Dumont from comment #14)
> Good news then.
> 
> On my side I only had some failures due to a faulty friend declaration in
> gnu-versioned-namespace mode in <format> for which I've submitted a patch:
> https://gcc.gnu.org/pipermail/libstdc++/2023-August/056560.html

Ah so probably that covers most of the libstdc++ cases - they do seem to be
format-related.

[Bug libstdc++/83077] sso-string @ gnu-versioned-namespace.

[fdumont at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83077

--- Comment #17 from François Dumont <fdumont at gcc dot gnu.org> ---
(In reply to Iain Sandoe from comment #15)
> 
> many of the c++ fails are of this form:
> 
> contracts-tmpl-spec1.C:(.text+0x6f): undefined reference to
> `handle_contract_violation(std::experimental::contract_violation const&)'

I'm surprised that my patch can have an impact on the C++ front end but I'll
give it a try. Did you first run those without my patches ?

I've never run the C++ tests. They can be run without a proper libstdc++ build,
no ?

[Bug libstdc++/83077] sso-string @ gnu-versioned-namespace.

[iains at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83077

--- Comment #18 from Iain Sandoe <iains at gcc dot gnu.org> ---
for changes to libstdc++ or the FE I usually run "make check-c++" which does
the library (plus the libgomp and itm deps) and the FE.

My guess is that the FE is referencing something that needs to have an inline
namespace added.

There are also some failing coroutine tests (because their scan strings need to
account for the inline namespace - I have a WIP patch for that, it's just
tedious editing of pattern matches).

[Bug libstdc++/83077] sso-string @ gnu-versioned-namespace.

[iains at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83077

--- Comment #19 from Iain Sandoe <iains at gcc dot gnu.org> ---
(In reply to François Dumont from comment #17)
> (In reply to Iain Sandoe from comment #15)

Your proposed patch for the friend issue does fix the libstdc++ cases for my
Darwin patchset.

> > many of the c++ fails are of this form:
> > 
> > contracts-tmpl-spec1.C:(.text+0x6f): undefined reference to
> > `handle_contract_violation(std::experimental::contract_violation const&)'
> 
> I'm surprised that my patch can have an impact on the C++ front end but I'll
> give it a try. 

I do not think it's affecting the FE as such - but rather that some of the
tests include stdc++ headers (I try to avoid it in the testsuite, but obviously
things like coroutines cannot avoid it)

[Bug libstdc++/83077] sso-string @ gnu-versioned-namespace.

[fdumont at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83077

--- Comment #20 from François Dumont <fdumont at gcc dot gnu.org> ---
I run make check-c++ before and after my patch and I see no regression. I even
have less failures with the patch even if I haven't check yet why.

So I think the patch is quite safe, just waiting for validation now.