Re: [PATCH 2/2] BFD error message suppression test case

[Kevin Buettner <kevinb@redhat.com>]

On Tue, 6 Sep 2022 10:29:08 +0000
Lancelot SIX <lsix@lancelotsix.com> wrote:

> When applying this patch, I have:
> 
>     Applying: BFD error message suppression test case
>     [...]/rebase-apply/patch:269: trailing whitespace.
>     # in the dynamic linker/loader due to using a mangled shared object. 
>     warning: 1 line adds whitespace errors.

Thanks.  I've fixed this in my local sources.

> Also, on my systems (ubuntu 22.04 and ubuntu 20.04), the test fails:
> 
>     (gdb) catch load bfd-errors-lib
>     Catchpoint 1 (load)
>     (gdb) PASS: gdb.base/bfd-errors.exp: catch load bfd-errors-lib
>     run 
>     Starting program: .../gdb/testsuite/outputs/gdb.base/bfd-errors/bfd-errors-main 
>     .../gdb/testsuite/outputs/gdb.base/bfd-errors/bfd-errors-main: symbol lookup error: .../gdb/testsuite/outputs/gdb.base/bfd-errors/bfd-errors-main: undefined symbol: foo4
>     [Inferior 1 (process 19165) exited with code 0177]
>     (gdb) FAIL: gdb.base/bfd-errors.exp: run to catchpoint (the program exited)
>     PASS: gdb.base/bfd-errors.exp: consolidated bfd errors
>     FAIL: gdb.base/bfd-errors.exp: print all unique bfd error

I've reproduced this on Ubuntu 22.04.

I was concerned that setting a catchpoint on the shared library load
might be fragile.  Sadly, it seems that it is.  I was expecting it to break
differently than what we're seeing on Ubuntu though.

I've perhaps come up with a different approach using 'add-symbol-file'
which I'll (try to) put into a v2 series.  If it works, this will take
the dynamic linker entirely out of the picture and also won't rely on
a load actually being caught.

> I also have minor comments inlined in the patch.

Thanks - I'll make those adjustments in the v2 series too.

Kevin

P.S. Thanks, too, for your comments on part 1 of this series.