From: Andrew Burgess <aburgess@redhat.com>
To: "Sébastien Darche" <sdarche@efficios.com>, gdb-patches@sourceware.org
Cc: "Sébastien Darche" <sdarche@efficios.com>
Subject: Re: [PATCH v2] gdb/breakpoint: remove assertion when printing internal breakpoints without locspecs
Date: Wed, 10 Jun 2026 16:16:56 +0100 [thread overview]
Message-ID: <877bo6qvbr.fsf@redhat.com> (raw)
In-Reply-To: <20260610141509.12950-1-sdarche@efficios.com>
Thanks, this looks great. I have one minor nit, no need to repost though...
Sébastien Darche <sdarche@efficios.com> writes:
> When debugging gdb with `set debug breakpoint on`, gdb would crash if
> the inferior happened to unload (through dlclose ()) a shared library
> that contains jit debugging symbols:
>
> ../../gdb/breakpoint.c:6473: internal-error: print_breakpoint_location:
> Assertion `b->locspec != nullptr || (!user_breakpoint_p (b) && (b->type
> == bp_shlib_event || b->type == bp_thread_event))' failed.
> A problem internal to GDB has been detected [...]
>
> This assertion was added in commit 5770f68 ("gdb: handle empty locspec
> when printing breakpoints"). The assumption for this assert is that the
> user may not call any (gdb-)debugging command between the time a shared
> library is marked (bp_location->shlib_disabled is set) and the
> breakpoint is actually removed. This would be true if the user called
> `maint info breakpoints` as explained in the original commit message.
>
> However, if `set debug breakpoint on` is called, gdb prints debugging
> info as the shared library is being unloaded. The ~jiter_objfile_data
> dtor deletes the breakpoint, which calls remove_breakpoint_1 on the
> bp_location in the recently unloaded shared object. Since `debug
> breakpoint` is enabled, we go through print_breakpoint_location with the
> bp_location's shlib_disabled flag set, which results in the assert
> above.
>
> This commit removes the assertion as it is not true in all cases. I've
> also included a minimal test case that loads a shared library with a
> __jit_debug_register_code symbol. The test fails with the assert.
>
> Change-Id: Ia19c84f194a6f0c10315548c7f423bfd86ef0266
> ---
> gdb/breakpoint.c | 17 -------
> gdb/testsuite/gdb.base/jit-unload-solib.c | 21 ++++++++
> gdb/testsuite/gdb.base/jit-unload.c | 45 ++++++++++++++++
> gdb/testsuite/gdb.base/jit-unload.exp | 62 +++++++++++++++++++++++
> 4 files changed, 128 insertions(+), 17 deletions(-)
> create mode 100644 gdb/testsuite/gdb.base/jit-unload-solib.c
> create mode 100644 gdb/testsuite/gdb.base/jit-unload.c
> create mode 100644 gdb/testsuite/gdb.base/jit-unload.exp
>
> diff --git a/gdb/breakpoint.c b/gdb/breakpoint.c
> index 101dc57ee6b..85c4ffc3171 100644
> --- a/gdb/breakpoint.c
> +++ b/gdb/breakpoint.c
> @@ -6457,23 +6457,6 @@ print_breakpoint_location (const breakpoint *b, const bp_location *loc)
> }
> else
> {
> - /* Internal breakpoints don't have a locspec string, but can become
> - pending if the shared library the breakpoint is in is unloaded.
> - For most internal breakpoint types though, after unloading the
> - shared library, the breakpoint will be deleted and never recreated
> - (see internal_breakpoint::re_set). But for two internal
> - breakpoint types bp_shlib_event and bp_thread_event this is not
> - true. Usually we don't expect the libraries that contain these
> - breakpoints to ever be unloaded, but a buggy inferior might do
> - such a thing, in which case GDB should be prepared to handle this
> - case.
> -
> - If these two breakpoint types become pending then there will be no
> - locspec string. */
> - gdb_assert (b->locspec != nullptr
> - || (!user_breakpoint_p (b)
> - && (b->type == bp_shlib_event
> - || b->type == bp_thread_event)));
> const char *locspec_str
> = (b->locspec != nullptr ? b->locspec->to_string () : "");
> uiout->field_string ("pending", locspec_str);
> diff --git a/gdb/testsuite/gdb.base/jit-unload-solib.c b/gdb/testsuite/gdb.base/jit-unload-solib.c
> new file mode 100644
> index 00000000000..3cf2c580320
> --- /dev/null
> +++ b/gdb/testsuite/gdb.base/jit-unload-solib.c
> @@ -0,0 +1,21 @@
> +/* This test program is part of GDB, the GNU debugger.
> +
> + Copyright 2026 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/>. */
> +
> +/* This simulates a JIT library. */
> +
> +
> +#include "jit-protocol.h"
> diff --git a/gdb/testsuite/gdb.base/jit-unload.c b/gdb/testsuite/gdb.base/jit-unload.c
> new file mode 100644
> index 00000000000..92dfa16c055
> --- /dev/null
> +++ b/gdb/testsuite/gdb.base/jit-unload.c
> @@ -0,0 +1,45 @@
> +/* This test program is part of GDB, the GNU debugger.
> +
> + Copyright 2026 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/>. */
> +
> +/* Simulate loading of a library JIT code. */
> +
> +#include <stdio.h>
> +#include <stdlib.h>
> +
> +#ifdef __WIN32__
> +#include <windows.h>
> +#define dlopen(name, mode) LoadLibrary (TEXT (name))
> +#define dlclose(handle) FreeLibrary (handle)
There's no definition of dlerror() here. Similar tests do:
#define dlerror() "an error occurred"
If you're happy to add that here then:
Approved-By: Andrew Burgess <aburgess@redhat.com>
Thanks,
Andrew
> +#else
> +#include <dlfcn.h>
> +#endif
> +
> +
> +int
> +main (void)
> +{
> + void *handle = dlopen (SHLIB_NAME, RTLD_NOW);
> +
> + if (handle == NULL)
> + {
> + fprintf (stderr, "%s\n", dlerror ());
> + exit (1);
> + }
> +
> + dlclose (handle);
> + return 0;
> +}
> diff --git a/gdb/testsuite/gdb.base/jit-unload.exp b/gdb/testsuite/gdb.base/jit-unload.exp
> new file mode 100644
> index 00000000000..a5f9adbd76c
> --- /dev/null
> +++ b/gdb/testsuite/gdb.base/jit-unload.exp
> @@ -0,0 +1,62 @@
> +# Copyright 2026 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/>.
> +
> +# This test checks that gdb does not assert when unloading a shared library
> +# that defines a __jit_debug_register_code if `set debug breakpoint on` is set.
> +
> +require allow_shlib_tests
> +
> +standard_testfile .c -solib.c
> +set lib_testfile ${testfile}-lib
> +set lib_binfile [standard_output_file $lib_testfile]
> +
> +if { [build_executable "build shlib" $lib_testfile $srcfile2 \
> + {debug shlib}] == -1 } {
> + return
> +}
> +
> +set lib_testfile_on_target [gdb_download_shlib $lib_binfile]
> +
> +set options [list \
> + debug \
> + shlib_load \
> + additional_flags=-DSHLIB_NAME=\"${lib_testfile_on_target}\"]
> +
> +if { [prepare_for_testing "build main file" $testfile $srcfile \
> + $options] == -1 } {
> + return
> +}
> +
> +if { ![runto_main] } {
> + return
> +}
> +
> +gdb_test_no_output "set debug breakpoint on"
> +
> +set saw_exit false
> +gdb_test_multiple "continue" unload {
> + -re "^\\\[Inferior $decimal \[^\r\n\]+ exited normally\\\]\r\n" {
> + set saw_exit true
> + exp_continue
> + }
> +
> + -re "^$gdb_prompt $" {
> + gdb_assert ${saw_exit} $gdb_test_name
> + }
> +
> + -re "^\[^\r\n\]*\r\n" {
> + exp_continue
> + }
> +}
> --
> 2.54.0
next prev parent reply other threads:[~2026-06-10 15:17 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-06-10 14:15 Sébastien Darche
2026-06-10 15:16 ` Andrew Burgess [this message]
2026-06-10 15:37 ` Sébastien Darche
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=877bo6qvbr.fsf@redhat.com \
--to=aburgess@redhat.com \
--cc=gdb-patches@sourceware.org \
--cc=sdarche@efficios.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox