[gcc-10-branch] libiberty: Support the new ("v0") mangling scheme in rust-demangle

[gcc-10-branch] libiberty: Support the new ("v0") mangling scheme in rust-demangle

[Amanieu d'Antras via Gdb-patches]

Rust is in the process of migrating to a new symbol mangling scheme called "v0".
Details can be found in the tracking issue:
https://github.com/rust-lang/rust/issues/60705

Demangler support for v0 has already been upstreamed in GCC's libiberty and is
available in gdb-binutils master. However we would like the demangler to be
backported to GDB 10 so that debugging experience does not regress when we
switch to generating v0 symbol names by default.

The specific GCC commit that contains the v0 Rust demangler is:
commit 84096498a7bd788599d4a7ca63543fc7c297645e
    libiberty: Support the new ("v0") mangling scheme in rust-demangle

The patch applies cleanly to gdb-10-branch.

Re: [gcc-10-branch] libiberty: Support the new ("v0") mangling scheme in rust-demangle

[Tom Tromey]

Amanieu> Demangler support for v0 has already been upstreamed in GCC's libiberty and is
Amanieu> available in gdb-binutils master. However we would like the demangler to be
Amanieu> backported to GDB 10 so that debugging experience does not regress when we
Amanieu> switch to generating v0 symbol names by default.

I don't know whether we plan to do another release from the GDB 10
branch.  Maybe Joel could say.

Also, is there a way to test that this works correctly?  I'd imagine
there just has to be a flag day sometime and then we find out?  Or has
someone built Rust and tried it with GDB trunk?

Anyway, it's fine by me to do this.  If there's no objections I can do
it sometime.  You may want to ping after a week or so in case I forget.

thanks,
Tom

Re: [gcc-10-branch] libiberty: Support the new ("v0") mangling scheme in rust-demangle

[Simon Marchi via Gdb-patches]

On 2021-01-12 8:02 p.m., Tom Tromey wrote:
> I don't know whether we plan to do another release from the GDB 10
> branch.  Maybe Joel could say.

GDB 10.2 hasn't been released, I suppose that one will be released.

Simon

Re: [gcc-10-branch] libiberty: Support the new ("v0") mangling scheme in rust-demangle

[Amanieu d'Antras via Gdb-patches]

> Also, is there a way to test that this works correctly?  I'd imagine
> there just has to be a flag day sometime and then we find out?  Or has
> someone built Rust and tried it with GDB trunk?

rustc can already generate v0 symbols with the
-Zsymbol-mangling-version=v0 flag. We're just going to eventually be
switching this on by default.

I just tested a small program built with that flag and gdb trunk can
successfully demangle the symbols while gdb 10 can't.

Re: [gcc-10-branch] libiberty: Support the new ("v0") mangling scheme in rust-demangle

[Tom Tromey]

Amanieu> rustc can already generate v0 symbols with the
Amanieu> -Zsymbol-mangling-version=v0 flag. We're just going to eventually be
Amanieu> switching this on by default.

Amanieu> I just tested a small program built with that flag and gdb trunk can
Amanieu> successfully demangle the symbols while gdb 10 can't.

Thanks, I ran the gdb test suite using nightly and that flag, and it
worked fine.

Tom

Re: [gcc-10-branch] libiberty: Support the new ("v0") mangling scheme in rust-demangle

[Tom Tromey]

Tom> Anyway, it's fine by me to do this.  If there's no objections I can do
Tom> it sometime.  You may want to ping after a week or so in case I forget.

I checked it in on the gdb 10 branch just now.  I've also filed a bug
noting the backport, per our usual procedure:

https://sourceware.org/bugzilla/show_bug.cgi?id=27194

Tom