From: Andrew Burgess <aburgess@redhat.com>
To: Guinevere Larsen <guinevere@redhat.com>, gdb-patches@sourceware.org
Cc: Guinevere Larsen <guinevere@redhat.com>,
Jan Kratochvil <jan@jankratochvil.net>,
Thiago Jung Bauermann <thiago.bauermann@linaro.org>
Subject: Re: [PATCH v8 5/5] gdb/testsuite: Test for a backtrace through object without debuginfo
Date: Thu, 16 Jan 2025 14:37:34 +0000 [thread overview]
Message-ID: <877c6uq0y9.fsf@redhat.com> (raw)
In-Reply-To: <20241210195115.3046370-6-guinevere@redhat.com>
Guinevere Larsen <guinevere@redhat.com> writes:
> Fedora has been carrying this test since back in the Project Archer
> days. A change back then caused GDB to stop being able to backtrace when
> only some of the object files had debug information. Even though the
> changed code never seems to have made its way into the main GDB project,
> I think it makes sense to bring the test along to ensure something like
> this doesn't pass unnoticed.
>
> Co-Authored-By: Jan Kratochvil <jan@jankratochvil.net>
> Reviewed-by: Thiago Jung Bauermann <thiago.bauermann@linaro.org>
> ---
> .../backtrace-through-cu-nodebug-caller.c | 28 ++++++
> .../backtrace-through-cu-nodebug-main.c | 32 +++++++
> .../gdb.base/backtrace-through-cu-nodebug.exp | 95 +++++++++++++++++++
> 3 files changed, 155 insertions(+)
> create mode 100644 gdb/testsuite/gdb.base/backtrace-through-cu-nodebug-caller.c
> create mode 100644 gdb/testsuite/gdb.base/backtrace-through-cu-nodebug-main.c
> create mode 100644 gdb/testsuite/gdb.base/backtrace-through-cu-nodebug.exp
>
> diff --git a/gdb/testsuite/gdb.base/backtrace-through-cu-nodebug-caller.c b/gdb/testsuite/gdb.base/backtrace-through-cu-nodebug-caller.c
> new file mode 100644
> index 00000000000..3a63d72a468
> --- /dev/null
> +++ b/gdb/testsuite/gdb.base/backtrace-through-cu-nodebug-caller.c
> @@ -0,0 +1,28 @@
> +/* This testcase is part of GDB, the GNU debugger.
> +
> + Copyright 2005-2024 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/>. */
> +
> +typedef int (*callback_t) (void);
> +
> +int
> +caller (callback_t callback)
> +{
> + /* Ensure some frame content to push away the return address. */
> + volatile const long one = 1;
> +
> + /* Modify the return value to prevent any tail-call optimization. */
> + return (*callback) () - one;
> +}
> diff --git a/gdb/testsuite/gdb.base/backtrace-through-cu-nodebug-main.c b/gdb/testsuite/gdb.base/backtrace-through-cu-nodebug-main.c
> new file mode 100644
> index 00000000000..3e7ac57a166
> --- /dev/null
> +++ b/gdb/testsuite/gdb.base/backtrace-through-cu-nodebug-main.c
> @@ -0,0 +1,32 @@
> +/* This testcase is part of GDB, the GNU debugger.
> +
> + Copyright 2005-2024 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/>. */
> +
> +typedef int (*callback_t) (void);
> +
> +extern int caller (callback_t callback);
> +
> +int
> +callback (void)
> +{
> + return 1;
> +}
> +
> +int
> +main (void)
> +{
> + return caller (callback);
> +}
> diff --git a/gdb/testsuite/gdb.base/backtrace-through-cu-nodebug.exp b/gdb/testsuite/gdb.base/backtrace-through-cu-nodebug.exp
> new file mode 100644
> index 00000000000..c0940b406a8
> --- /dev/null
> +++ b/gdb/testsuite/gdb.base/backtrace-through-cu-nodebug.exp
> @@ -0,0 +1,95 @@
> +# Copyright 2010-2024 Free Software Foundation, Inc.
Remember to update the copyright year throughout.
> +
> +# 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 that GDB can generate accurate backtraces even if some of the stack
> +# trace goes through a function with no debug information.
> +
> +standard_testfile -caller.c -main.c
> +set objmainfile ${testfile}-main.o
> +set objcallerfile ${testfile}-caller.o
> +
> +# recompile the inferior with or without CFI information, then run the
> +# inferior until the point where the important test starts
> +# returns TRUE on an ERROR.
Needs converting to two sentences ('.' after 'starts'). Plus capital
'R' for each sentence.
I wonder if it would be better to return TRUE on success, and FALSE on
error, because .... see below ...
> +proc prepare_test {has_cfi} {
> + global srcdir subdir srcfile srcfile2 objmainfile objcallerfile binfile
> + if {$has_cfi} {
> + set extension "cfi"
> + if {[gdb_compile "${srcdir}/${subdir}/${srcfile}" \
> + "${srcdir}/${subdir}/${objcallerfile}" \
> + object [list {additional_flags=-fomit-frame-pointer \
> + -funwind-tables -fasynchronous-unwind-tables}]] != "" } {
> + untested "couldn't compile with cfi"
> + return true
> + }
> + } else {
> + set extension "no-cfi"
> + if {[gdb_compile "${srcdir}/${subdir}/${srcfile}" \
> + "${srcdir}/${subdir}/${objcallerfile}" \
> + object [list {additional_flags=-fomit-frame-pointer \
> + -fno-unwind-tables \
> + -fno-asynchronous-unwind-tables}]] != "" } {
> + untested "couldn't compile without cfi"
> + return true
> + }
> + }
> + if {[gdb_compile [list "${srcdir}/${subdir}/${objmainfile}" \
> + "${srcdir}/${subdir}/${objcallerfile}"] \
> + "${binfile}-${extension}" binfile {}] != ""} {
> + untested "couldn't link object files"
> + return true
> + }
> +
> + clean_restart "$binfile-${extension}"
> +
> + with_test_prefix "${extension}" {
> +
> + if ![runto callback] then {
> + fail "has_cfi=$has_cfi: Can't run to callback"
> + return true
> + }
> + gdb_test_no_output "maint frame-unwinder disable ARCH"
> + return false
> + }
> +}
> +
> +if {[gdb_compile "${srcdir}/${subdir}/${srcfile2}" \
> + "${srcdir}/${subdir}/${objmainfile}" \
> + object {debug}] != "" } {
> + untested "couldn't compile main file"
> + return
> +}
> +
> +if { [prepare_test false] } {
> + untested ${testfile}.exp
> +} else {
> + gdb_test "bt" "Required frame unwinder may have been disabled.*" \
> + "verify unwind fail without CFI"
> +}
When something goes wrong with the prepare_test call we've already
emitted an 'untested' or 'fail' message, thus I think we can just write:
if { [prepare_test false] } {
... perform the test ...
}
Of course, this assumes that the true/false return value for
prepare_test have been switched. Currently you'd write:
if { ![prepare_test false] } {
... perform the test ...
}
which just seems weird.
Also, the test here says "verify unwind fail without CFI". And we _do_
check for the error message about disabled unwinders. But if the `bt`
worked this test would still pass just fine. We should probably be
checking that we only print frame #0. Something like:
gdb_test "bt" \
[multi_line \
"\[^\r\n\]+Required frame unwinder may have been disabled, \[^\r\n\]+" \
"#0\\s+callback \\(\\) \[^\r\n\]+"] \
"verify unwind fail without CFI"
should do the job I think.
> +
> +if { [prepare_test true] } {
> + untested ${testfile}.exp
> +} else {
Same suggestion here about avoiding the extra 'untested' call.
> + if { [istarget "arm*-*-*"] } {
> + setup_kfail backtrace/31950 *-*-*
> + }
> + set text {[^\r\n]+}
> + # #0 callback () at ...
> + # #1 0x00000000004004e9 in caller ()
> + # #2 0x00000000004004cd in main () at ...
> + gdb_test "bt" \
> + "#0 +callback $text\r\n#1 $text in caller $text\r\n#2 $text in main $text" \
> + "verify unwinding works for CFI without DIEs"
The test name here seems weird. Maybe: 'verify unwinding works for CUs
without CFI' would be better?
OK with these fixes.
Approved-By: Andrew Burgess <aburgess@redhat.com>
Thanks,
Andrew
> +}
> --
> 2.47.0
next prev parent reply other threads:[~2025-01-16 14:38 UTC|newest]
Thread overview: 24+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-12-10 19:51 [PATCH v8 0/5] Modernize frame unwinders and add disable feature Guinevere Larsen
2024-12-10 19:51 ` [PATCH v8 1/5] gdb: make gdbarch store a vector of frame unwinders Guinevere Larsen
2025-01-14 14:28 ` Andrew Burgess
2025-01-14 20:34 ` Guinevere Larsen
2024-12-10 19:51 ` [PATCH v8 2/5] gdb: add "unwinder class" to " Guinevere Larsen
2025-01-14 15:28 ` Andrew Burgess
2024-12-10 19:51 ` [PATCH v8 3/5] gdb: Migrate frame unwinders to use C++ classes Guinevere Larsen
2025-01-14 17:13 ` Andrew Burgess
2024-12-10 19:51 ` [PATCH v8 4/5] gdb: introduce ability to disable frame unwinders Guinevere Larsen
2025-01-16 12:06 ` Andrew Burgess
2025-01-17 12:40 ` Guinevere Larsen
2025-01-17 13:55 ` Andrew Burgess
2025-01-17 14:47 ` Guinevere Larsen
2025-01-16 16:22 ` Andrew Burgess
2024-12-10 19:51 ` [PATCH v8 5/5] gdb/testsuite: Test for a backtrace through object without debuginfo Guinevere Larsen
2025-01-16 14:37 ` Andrew Burgess [this message]
2025-01-16 18:42 ` Guinevere Larsen
2025-01-17 13:58 ` Andrew Burgess
2025-01-18 8:07 ` Tom de Vries
2025-01-20 12:26 ` [PATCH] gdb/testsuite: Fix file location for gdb.base/backtrace-through-cu-nodebug Guinevere Larsen
2025-01-20 12:46 ` Tom de Vries
2025-01-20 12:48 ` Guinevere Larsen
2025-01-07 12:11 ` [PING][PATCH v8 0/5] Modernize frame unwinders and add disable feature Guinevere Larsen
2025-01-17 14:49 ` [PATCH " Guinevere Larsen
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=877c6uq0y9.fsf@redhat.com \
--to=aburgess@redhat.com \
--cc=gdb-patches@sourceware.org \
--cc=guinevere@redhat.com \
--cc=jan@jankratochvil.net \
--cc=thiago.bauermann@linaro.org \
/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