diff options
| author | Simon Marchi <simon.marchi@polymtl.ca> | 2022-12-13 22:34:39 -0500 |
|---|---|---|
| committer | Simon Marchi <simon.marchi@efficios.com> | 2023-01-20 14:48:57 -0500 |
| commit | 836a8d3710462596bb4184617999a9507adc3629 (patch) | |
| tree | da76674260e0c55130bfe6931918cfd6a6b8e643 /gdb/Makefile.in | |
| parent | bc2cbe815bdbac3bd027bf4acc94c554c29b0189 (diff) | |
| download | gdb-836a8d3710462596bb4184617999a9507adc3629.zip gdb-836a8d3710462596bb4184617999a9507adc3629.tar.gz gdb-836a8d3710462596bb4184617999a9507adc3629.tar.bz2 | |
gdb: make user-created frames reinflatable
This patch teaches frame_info_ptr to reinflate user-created frames
(frames created through create_new_frame, with the "select-frame view"
command).
Before this patch, frame_info_ptr doesn't support reinflating
user-created frames, because it currently reinflates by getting the
current target frame (for frame 0) or frame_find_by_id (for other
frames). To reinflate a user-created frame, we need to call
create_new_frame, to make it lookup an existing user-created frame, or
otherwise create one.
So, in prepare_reinflate, get the frame id even if the frame has level
0, if it is user-created. In reinflate, if the saved frame id is user
create it, call create_new_frame.
In order to test this, I initially enhanced the gdb.base/frame-view.exp
test added by the previous patch by setting a pretty-printer for the
type of the function parameters, in which we do an inferior call. This
causes print_frame_args to not reinflate its frame (which is a
user-created one) properly. On one machine (my Arch Linux one), it
properly catches the bug, as the frame is not correctly restored after
printing the first parameter, so it messes up the second parameter:
frame
#0 baz (z1=hahaha, z2=<error reading variable: frame address is not available.>) at /home/simark/src/binutils-gdb/gdb/testsuite/gdb.base/frame-view.c:40
40 return z1.m + z2.n;
(gdb) FAIL: gdb.base/frame-view.exp: with_pretty_printer=true: frame
frame
#0 baz (z1=hahaha, z2=<error reading variable: frame address is not available.>) at /home/simark/src/binutils-gdb/gdb/testsuite/gdb.base/frame-view.c:40
40 return z1.m + z2.n;
(gdb) FAIL: gdb.base/frame-view.exp: with_pretty_printer=true: frame again
However, on another machine (my Ubuntu 22.04 one), it just passes fine,
without the appropriate fix. I then thought about writing a selftest
for that, it's more reliable. I left the gdb.base/frame-view.exp pretty
printer test there, it's already written, and we never know, it might
catch some unrelated issue some day.
Change-Id: I5849baf77991fc67a15bfce4b5e865a97265b386
Reviewed-By: Bruno Larsen <blarsen@redhat.com>
Diffstat (limited to 'gdb/Makefile.in')
| -rw-r--r-- | gdb/Makefile.in | 1 |
1 files changed, 1 insertions, 0 deletions
diff --git a/gdb/Makefile.in b/gdb/Makefile.in index cce85b8..c3711a0 100644 --- a/gdb/Makefile.in +++ b/gdb/Makefile.in @@ -458,6 +458,7 @@ SELFTESTS_SRCS = \ unittests/environ-selftests.c \ unittests/filtered_iterator-selftests.c \ unittests/format_pieces-selftests.c \ + unittests/frame_info_ptr-selftests.c \ unittests/function-view-selftests.c \ unittests/gdb_tilde_expand-selftests.c \ unittests/gmp-utils-selftests.c \ |