Re: [PATCH v3] nptl_db: Support different libpthread/ld.so load orders (bug 27744)

["H.J. Lu" <hjl.tools@gmail.com>]

On Wed, Apr 21, 2021 at 6:04 AM Florian Weimer via Libc-alpha
<libc-alpha@sourceware.org> wrote:
>
> * Florian Weimer via Libc-alpha:
>
> > libthread_db is loaded once GDB encounters libpthread, and at this
> > point, ld.so may not have been processed by GDB yet. As a result,
> > _rtld_global cannot be accessed by regular means from libthread_db.
> > To make this work until GDB can be fixed, acess _rtld_global through
> > a pointer stored in libpthread.
> >
> > The new test does not reproduce bug 27744 with
> > --disable-hardcoded-path-in-tests, but is still a valid smoke test.
> > With --enable-hardcoded-path-in-tests, it is necessary to avoid
> > add-symbol-file because this can tickle a GDB bug.
> >
> > Fixes commit 1daccf403b1bd86370eb94edca794dc106d02039 ("nptl: Move
> > stack list variables into _rtld_global").
> >
> > ---
> > v3: Mark test as UNSUPPORTED if launching gdb fails with ENOENT.
> >     This is the version I'd like to commit soon.
>
> I've pushed this version given that Pedro acknowledged v2.
>

I got

FAIL: nptl/tst-pthread-gdb-attach
FAIL: nptl/tst-pthread-gdb-attach-static

when testing i686 build on Fedora 33:

[New LWP 1329590]
Trying host libthread_db library:
/export/build/gnu/tools-build/glibc-32bit-cet-gitlab/build-i686-linux/nptl_db/libthread_db.so.1.
dlopen failed: /export/build/gnu/tools-build/glibc-32bit-cet-gitlab/build-i686-linux/nptl_db/libthread_db.so.1:
wrong ELF class: ELFCLASS32.
thread_db_load_search returning 0

Is it possible to skip this when gdb and libthread_db.so.1 have different
ELC classes?

-- 
H.J.