* [PATCH v5] gdb/debuginfod: Ctrl-C asks to cancel further downloads
@ 2024-10-09 22:07 Aaron Merey
2024-10-14 11:15 ` Schimpe, Christina
0 siblings, 1 reply; 2+ messages in thread
From: Aaron Merey @ 2024-10-09 22:07 UTC (permalink / raw)
To: gdb-patches; +Cc: Aaron Merey, Martin Liška, Tom de Vries
Currently gdb only allows for debuginfod downloads to be cancelled one at
a time via Ctrl-C. This can be a burden if one wishes to cancel a large
series of downloads. Additionally there can also be ambiguity between whether
Ctrl-C during a download was intended to cancel a download or interrupt the
inferior.
This patch addresses these issues by adding a new debuginfod setting and
changing the behavior of Ctrl-C during a download.
Add a new command "set debuginfod cancel one/all/ask", where:
- "one" means Ctrl-C cancels one download,
- "all" means Ctrl-C cancels all further downloads, and
- "ask" means Ctrl-C asks whether to cancel all further downloads. A "yes"
implies "set debuginfod cancel all", and a "no" implies "set debuginfod
cancel one", so the question is only asked once.
Note that the behaviour as it was before this patch is equivalent to
"set debuginfod cancel one". Instead, the new default is "set debuginfod
cancel ask". Note that cancelling all further downloads implies "set
debuginfod enabled off".
A single Ctrl-C during downloading now sets the quit_flag and proceeds with
all downloads. If the inferior has the terminal, then a second Ctrl-C during
downloading triggers a query asking whether to cancel the download or
interrupt the inferior. If the user wishes to cancel the download then
the setting of 'set debuginfod cancel' determines whether one or all
downloads are cancelled. In the case of "set debuginfod cancel ask",
there will be another query at this point asking whether to cancel one
or all downloads.
If the inferior does not have the terminal, then a second Ctrl-C during
downloading simply cancels the download according to the setting of
"set debuginfod cancel". In this case there is no query asking whether
to interrupt the inferior or cancel a download.
Example session where inferior has terminal:
(gdb) run
[...]
Downloading separate debug info for /lib64/libxyz.so
[### ]^C^C
Cancel the current download?
If no, then Ctrl-C will be sent to the target process. ([y] or n) y
Cancelling download of separate debug info for /lib64/libxyz.so...
Cancel further downloading for this session? (y or [n]) n
Downloading separate debug info for /lib64/libabcd.so
Example session where inferior does not have terminal:
(gdb) run
[...]
Downloading separate debug info for /lib64/libxyz.so
[### ]^C^C
Cancelling download of separate debug info for /lib64/libxyz.so...
Cancel further downloading for this session? (y or [n]) y
Debuginfod has been disabled.
To re-enable use the 'set debuginfod enabled' command.
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29582
Suggested-By: Martin Liška <mliska@suse.cz>
Co-Authored-By: Tom de Vries <tdevries@suse.de>
---
Eli has already approved the documentation parts of the patch.
This patch has previously been posted [1]. I'm reposting here in a
separate thread to make it more clear what the current up-to-date
version of this patch is.
Aaron
[1] https://sourceware.org/pipermail/gdb-patches/2024-June/209896.html
gdb/debuginfod-support.c | 174 +++++++++++++++++++++++++++++++++++++--
gdb/doc/gdb.texinfo | 21 +++++
2 files changed, 189 insertions(+), 6 deletions(-)
diff --git a/gdb/debuginfod-support.c b/gdb/debuginfod-support.c
index 841b6f2078c..28cedf6b7a8 100644
--- a/gdb/debuginfod-support.c
+++ b/gdb/debuginfod-support.c
@@ -37,6 +37,8 @@ static cmd_list_element *maint_show_debuginfod_cmdlist;
static const char debuginfod_on[] = "on";
static const char debuginfod_off[] = "off";
static const char debuginfod_ask[] = "ask";
+static const char debuginfod_one[] = "one";
+static const char debuginfod_all[] = "all";
static const char *debuginfod_enabled_enum[] =
{
@@ -46,6 +48,20 @@ static const char *debuginfod_enabled_enum[] =
nullptr
};
+/* Valid values for set debuginfod cancel command. */
+
+static const char *debuginfod_cancel_enum[] =
+{
+ debuginfod_one,
+ debuginfod_all,
+ debuginfod_ask,
+ nullptr
+};
+
+/* Value of debuginfod cancellation mode. */
+
+static const char *debuginfod_cancel = debuginfod_ask;
+
static const char *debuginfod_enabled =
#if defined(HAVE_LIBDEBUGINFOD)
debuginfod_ask;
@@ -108,9 +124,17 @@ debuginfod_section_query (const unsigned char *build_id,
struct user_data
{
user_data (const char *desc, const char *fname)
- : desc (desc), fname (fname)
- { }
+ : pass_quit_flag (false), desc (desc), fname (fname)
+ {
+ /* Record whether the current inferior owns the terminal. */
+ if (!target_terminal::is_ours ())
+ inf_had_term = true;
+ else
+ inf_had_term = false;
+ }
+ bool pass_quit_flag;
+ bool inf_had_term;
const char * const desc;
const char * const fname;
ui_out::progress_update progress;
@@ -155,10 +179,84 @@ progressfn (debuginfod_client *c, long cur, long total)
if (check_quit_flag ())
{
ui_file *outstream = get_unbuffered (gdb_stdout);
- gdb_printf (outstream, _("Cancelling download of %s %s...\n"),
- data->desc, styled_fname.c_str ());
- return 1;
- }
+
+ /* If a single Ctrl-C occurs during downloading, let it propagate to the
+ target. If more than one Ctrl-C occurs, ask whether to cancel the
+ current download or interrupt the target. If the download is
+ cancelled, the setting of debuginfod_cancel will determine whether
+ the current download is cancelled or debuginfod is disabled. */
+ if (!data->pass_quit_flag)
+ data->pass_quit_flag = true;
+ else
+ {
+ int resp = 1;
+ bool extra_nl = false;
+
+ if (data->inf_had_term)
+ {
+ /* If Ctrl-C occurs during the following prompts, catch the
+ exception to prevent unsafe early returns to gdb's main
+ event loop. During these prompts, Ctrl-C is equivalent to
+ answering 'y'. */
+ try
+ {
+ resp = yquery (_("Cancel the current download?\nIf no, "
+ "then Ctrl-C will be sent to the target "
+ "process. "));
+ }
+ catch (const gdb_exception &)
+ {
+ /* If the query doesn't complete, then we need an additional
+ newline to get "Cancelling download of..." printed on a
+ separate line. */
+ extra_nl = true;
+ }
+ }
+ if (resp)
+ {
+ if (extra_nl)
+ {
+ gdb_printf (outstream, "\n");
+ extra_nl = false;
+ }
+
+ gdb_printf (outstream, _("Cancelling download of %s %s...\n"),
+ data->desc, styled_fname.c_str ());
+ if (debuginfod_cancel == debuginfod_ask)
+ {
+ try
+ {
+ resp = nquery
+ (_("Cancel further downloading for this session? "));
+ }
+ catch (const gdb_exception &)
+ {
+ resp = 1;
+ extra_nl = true;
+ }
+
+ if (resp)
+ debuginfod_cancel = debuginfod_all;
+ else
+ debuginfod_cancel = debuginfod_one;
+ }
+ if (debuginfod_cancel == debuginfod_all)
+ {
+ if (extra_nl)
+ gdb_printf (outstream, "\n");
+
+ gdb_printf (outstream,
+ _("Interrupted -- debuginfod will be disabled.\n"
+ "To re-enable use the 'set debuginfod "
+ "enabled' command.\n"));
+ debuginfod_enabled = debuginfod_off;
+ }
+
+ data->pass_quit_flag = false;
+ return 1;
+ }
+ }
+ }
if (debuginfod_verbose == 0)
return 0;
@@ -334,6 +432,11 @@ debuginfod_source_query (const unsigned char *build_id,
build_id_len,
srcpath,
&dname));
+ if (data.pass_quit_flag)
+ set_quit_flag ();
+ if (data.inf_had_term && term_state.has_value ())
+ target_terminal::inferior ();
+
debuginfod_set_user_data (c, nullptr);
}
@@ -377,6 +480,11 @@ debuginfod_debuginfo_query (const unsigned char *build_id,
fd = scoped_fd (debuginfod_find_debuginfo (c, build_id, build_id_len,
&dname));
+ if (data.pass_quit_flag)
+ set_quit_flag ();
+ if (data.inf_had_term && term_state.has_value ())
+ target_terminal::inferior ();
+
debuginfod_set_user_data (c, nullptr);
}
@@ -420,6 +528,11 @@ debuginfod_exec_query (const unsigned char *build_id,
fd = scoped_fd (debuginfod_find_executable (c, build_id, build_id_len,
&dname));
+ if (data.pass_quit_flag)
+ set_quit_flag ();
+ if (data.inf_had_term && term_state.has_value ())
+ target_terminal::inferior ();
+
debuginfod_set_user_data (c, nullptr);
}
@@ -468,6 +581,11 @@ debuginfod_section_query (const unsigned char *build_id,
fd = scoped_fd (debuginfod_find_section (c, build_id, build_id_len,
section_name, &dname));
+ if (data.pass_quit_flag)
+ set_quit_flag ();
+ if (data.inf_had_term && term_state.has_value ())
+ target_terminal::inferior ();
+
debuginfod_set_user_data (c, nullptr);
}
@@ -516,6 +634,33 @@ show_debuginfod_enabled (ui_file *file, int from_tty, cmd_list_element *cmd,
"\"%s\".\n"), debuginfod_enabled);
}
+/* Set callback for "set debuginfod cancel". */
+
+static void
+set_debuginfod_cancel (const char *value)
+{
+ debuginfod_cancel = value;
+}
+
+/* Get callback for "set debuginfod cancel". */
+
+static const char *
+get_debuginfod_cancel ()
+{
+ return debuginfod_cancel;
+}
+
+/* Show callback for "set debuginfod cancel". */
+
+static void
+show_debuginfod_cancel (ui_file *file, int from_tty, cmd_list_element *cmd,
+ const char *value)
+{
+ gdb_printf (file,
+ _("Debuginfod cancellation mode is currently set to "
+ "\"%s\".\n"), debuginfod_cancel);
+}
+
/* Set callback for "set debuginfod urls". */
static void
@@ -620,6 +765,23 @@ When set to \"ask\", prompt whether to enable or disable debuginfod." ),
&set_debuginfod_prefix_list,
&show_debuginfod_prefix_list);
+ add_setshow_enum_cmd ("cancel", class_run, debuginfod_cancel_enum,
+ _("Set Ctrl-C behaviour for debuginfod."),
+ _("Show Ctrl-C behaviour for debuginfod."),
+ _("\
+When set to \'one\', pressing Ctrl-C twice cancels a single \
+download.\nWhen set to \'all\', pressing Ctrl-C twice cancels all further downloads.\n\
+When set to \'ask\', pressing Ctrl-C twice asks what to do.\nA single Ctrl-C during \
+downloading is passed to the target process being debugged.\nA second Ctrl-C \
+during downloading may raise a prompt asking whether to cancel the download or \
+send Ctrl-C to the target.\nIf the download is cancelled, then no Ctrl-C is \
+sent to the target."),
+ set_debuginfod_cancel,
+ get_debuginfod_cancel,
+ show_debuginfod_cancel,
+ &set_debuginfod_prefix_list,
+ &show_debuginfod_prefix_list);
+
/* set/show debuginfod urls */
add_setshow_string_noescape_cmd ("urls", class_run, _("\
Set the list of debuginfod server URLs."), _("\
diff --git a/gdb/doc/gdb.texinfo b/gdb/doc/gdb.texinfo
index 46ca62ec0c3..4a2f2f16b33 100644
--- a/gdb/doc/gdb.texinfo
+++ b/gdb/doc/gdb.texinfo
@@ -50502,6 +50502,27 @@ is set to @code{ask} for interactive sessions.
Display whether @code{debuginfod enabled} is set to @code{on}, @code{off} or
@code{ask}.
+@kindex set debuginfod cancel
+@anchor{set debuginfod cancel}
+@item set debuginfod cancel
+@itemx set debuginfod cancel one
+@cindex debuginfod @kbd{Ctrl-C} behaviour
+Pressing @kbd{Ctrl-C} twice during downloading will cancel the current
+download.
+
+@item set debuginfod cancel all
+Pressing @kbd{Ctrl-C} twice during downloading will cancel this and all
+further downloads.
+
+@item set debuginfod cancel ask
+Pressing @kbd{Ctrl-C} twice during downloading will cancel the current
+download and prompt whether to cancel all further downloads. By default,
+@code{debuginfod cancel} is set to @code{ask} for interactive sessions.
+
+@kindex show debuginfod cancel
+@item show debuginfod cancel
+Display the current setting of @code{debuginfod cancel}.
+
@kindex set debuginfod urls
@cindex configure debuginfod URLs
@item set debuginfod urls
--
2.46.2
^ permalink raw reply [flat|nested] 2+ messages in thread
* RE: [PATCH v5] gdb/debuginfod: Ctrl-C asks to cancel further downloads
2024-10-09 22:07 [PATCH v5] gdb/debuginfod: Ctrl-C asks to cancel further downloads Aaron Merey
@ 2024-10-14 11:15 ` Schimpe, Christina
0 siblings, 0 replies; 2+ messages in thread
From: Schimpe, Christina @ 2024-10-14 11:15 UTC (permalink / raw)
To: Aaron Merey, gdb-patches; +Cc: Martin Liška, Tom de Vries
Hi Aaron,
Thanks a lot for this patch. I really appreciate it.
Please see my feedback below.
> -----Original Message-----
> From: Aaron Merey <amerey@redhat.com>
> Sent: Thursday, October 10, 2024 12:08 AM
> To: gdb-patches@sourceware.org
> Cc: Aaron Merey <amerey@redhat.com>; Martin Liška <mliska@suse.cz>;
> Tom de Vries <tdevries@suse.de>
> Subject: [PATCH v5] gdb/debuginfod: Ctrl-C asks to cancel further downloads
>
> Currently gdb only allows for debuginfod downloads to be cancelled one at a
> time via Ctrl-C. This can be a burden if one wishes to cancel a large series of
> downloads. Additionally there can also be ambiguity between whether Ctrl-C
> during a download was intended to cancel a download or interrupt the
> inferior.
>
> This patch addresses these issues by adding a new debuginfod setting and
> changing the behavior of Ctrl-C during a download.
>
> Add a new command "set debuginfod cancel one/all/ask", where:
> - "one" means Ctrl-C cancels one download,
> - "all" means Ctrl-C cancels all further downloads, and
> - "ask" means Ctrl-C asks whether to cancel all further downloads. A "yes"
> implies "set debuginfod cancel all", and a "no" implies "set debuginfod
> cancel one", so the question is only asked once.
>
> Note that the behaviour as it was before this patch is equivalent to "set
> debuginfod cancel one". Instead, the new default is "set debuginfod cancel
> ask". Note that cancelling all further downloads implies "set debuginfod
> enabled off".
If I understand correctly, the behaviour is not exactly the same as before with
that setting configured. With your patch and the config
"set debuginfod cancel one", we see following output:
~~~
(gdb) set debuginfod cancel one
...
Downloading separate debug info for /.../libxyz.so
...
Cancel the current download?
If no, then Ctrl-C will be sent to the target process. ([y] or n) y
~~~
which is not the same as with current master:
~~~
Downloading separate debug info for /.../libxyz1.so
[######### ] 6% (1.89)
^CCancelling download of separate debug info for /.../libxyz1.so...
Downloading separate debug info for /.../libxyz2.so
~~~
With current master we automatically cancel the download after Ctrl-C.
Also with your patch, we have to send Ctrl-C twice, which was not necessary
before, too. Could you make that a bit more clear in the commit message?
And one last open I have without knowing how feasible it is:
Have you considered writing a test for this?
Best Regards,
Christina
Intel Deutschland GmbH
Registered Address: Am Campeon 10, 85579 Neubiberg, Germany
Tel: +49 89 99 8853-0, www.intel.de
Managing Directors: Sean Fennelly, Jeffrey Schneiderman, Tiffany Doon Silva
Chairperson of the Supervisory Board: Nicole Lau
Registered Office: Munich
Commercial Register: Amtsgericht Muenchen HRB 186928
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2024-10-14 11:15 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2024-10-09 22:07 [PATCH v5] gdb/debuginfod: Ctrl-C asks to cancel further downloads Aaron Merey
2024-10-14 11:15 ` Schimpe, Christina
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).