[PATCH 2/2] gdb/testsuite: Add gdb.base/memops-watchpoint.exp

[Thiago Jung Bauermann]

Thank you for your review!

Kevin Buettner <kevinb@redhat.com> writes:

> On Sat, 20 Apr 2024 18:33:07 -0300
> Thiago Jung Bauermann <thiago.bauermann@linaro.org> wrote:
>
>> diff --git a/gdb/testsuite/gdb.base/memops-watchpoint.exp
>> b/gdb/testsuite/gdb.base/memops-watchpoint.exp
>> new file mode 100644
>> index 000000000000..6fc84eb469c4
>> --- /dev/null
>> +++ b/gdb/testsuite/gdb.base/memops-watchpoint.exp
>> @@ -0,0 +1,83 @@
>> +# Copyright 2024 Free Software Foundation, Inc.
>> +
>> +# This program is free software; you can redistribute it and/or modify
>> +# it under the terms of the GNU General Public License as published by
>> +# the Free Software Foundation; either version 3 of the License, or
>> +# (at your option) any later version.
>> +#
>> +# This program is distributed in the hope that it will be useful,
>> +# but WITHOUT ANY WARRANTY; without even the implied warranty of
>> +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
>> +# GNU General Public License for more details.
>> +#
>> +# You should have received a copy of the GNU General Public License
>> +# along with this program.  If not, see <http://www.gnu.org/licenses/>.
>> +
>> +# Test a binary that uses standard libc memory operation functions.  They are
>> +# frequently optimized with specialized instructions, so make sure GDB behaves
>> +# correctly in their presence.
>> +
>> +# It's not possible to check in which libc function the watchpoint triggers
>> +# without its debug info.
>> +require libc_has_debug_info
>
> I'm wondering about the need for this requirement.  When I comment it
> out and run it on a machine without libc debuginfo, I do see 3 FAILs,
> but it seems to me that those could be turned into PASSes by changing
> the regular expressions for the "continue until..." tests.
>
> E.g. for the first one, with libc debuginfo, I see:
>
> continue
> Continuing.
>
> Hardware watchpoint 2: -location a[31]
>
> Old value = 101 'e'
> New value = 0 '\000'
> __memset_avx2_unaligned () at ../sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S:146
> 146		VMOVU	%VMM(0), (%rdi)
> (gdb) PASS: gdb.base/memops-watchpoint.exp: continue until memset watchpoint hits
>
> But, without libc debuginfo, the watchpoint still works:
>
> continue
> Continuing.
>
> Hardware watchpoint 2: -location a[31]
>
> Old value = 101 'e'
> New value = 0 '\000'
> 0x00007ffff7e3553a in __memset_avx2_unaligned () from /lib64/libc.so.6
> (gdb) FAIL: gdb.base/memops-watchpoint.exp: continue until memset watchpoint hits
>
> As stated earlier, this could be turned into a PASS by tweaking the RE.
>
> In both cases, we know that it's in a "memset" function.  (The presence
> of minimal symbols provides GDB with this information.)

I added the requirement because in my aarch64-linux system without libc6
debug info I get:

continue
Continuing.

Hardware watchpoint 2: -location a[28]

Old value = 104 'h'
New value = 0 '\000'
0x0000fffff7e90664 in ?? () from /lib/aarch64-linux-gnu/libc.so.6
(gdb) FAIL: gdb.base/memops-watchpoint.exp: continue until memset watchpoint hits

And I just tested removing libc6-dbg from my x86_64-linux laptop:

continue
Continuing.

Hardware watchpoint 2: -location a[28]

Old value = 104 'h'
New value = 0 '\000'
0x00007ffff7d8e05f in ?? () from /lib/x86_64-linux-gnu/libc.so.6
(gdb) FAIL: gdb.base/memops-watchpoint.exp: continue until memset watchpoint hits

So it depends on the system.

One alternative would be to not use the require statement and run the
test until the watchpoint hits, and have a case in gdb_test_multiple to
mark as UNRESOLVED if the function name is '??'.

--
Thiago