Re: [PATCH 1/2] gdb: add set/show commands for managing debuginfod

[Keith Seitz <keiths@redhat.com>]

Hi,

Thank you for your perseverance with this patch. This is really
shaping up as a nice, user-friendly addition. It's also very
"gdb-ish."

I have just a few things. Nothing major.

On 10/18/21 16:01, Aaron Merey wrote:

>    This GDB supports auto-downloading debuginfo from the following URLs:
>    <URL1> <URL2> ...
>    Enable debuginfod? (y or [n]) y
>    Debuginfod has been enabled.
>    To make this setting permanent, add 'set debuginfod on' to .gdbinit.

Very nice!

> Also add 'set/show debuginfod-urls' command. Accepts a string of
> space-separated debuginfod server URLs to be queried. The default
> value is copied from $DEBUGINFOD_URLS.

At first I thought this space-separated list as a bit odd, but I see that
this is exactly the same format as the environment variable DEBUGINFOD_URLS.
I certainly appreciate the consistency.

> Finally add 'set debug debuginfod" command to control whether
> debuginfod-related debugging output is displayed. This debugging
> output is displayed by default.
> 
>    (gdb) run
>    Starting program: /bin/sleep 5
>    Download failed: No route to host.  Continuing without debug info for /lib64/libc.so.6.

Here I have a question. From the patch, it appears that the debugging output
consists entirely of error messages? Am I alone thinking we should always print these
for the user?

> diff --git a/gdb/debuginfod-support.c b/gdb/debuginfod-support.c
> index 2d626e335a0..1a4a6e1dde0 100644
> --- a/gdb/debuginfod-support.c
> +++ b/gdb/debuginfod-support.c
> @@ -68,6 +69,63 @@ struct debuginfod_client_deleter
>   using debuginfod_client_up
>     = std::unique_ptr<debuginfod_client, debuginfod_client_deleter>;
>   
> +static const char debuginfod_on[] = "on";
> +static const char debuginfod_off[] = "off";
> +static const char debuginfod_ask[] = "ask";
> +static const char *const debuginfod_cmd_names[] =
> +{
> +  debuginfod_on,
> +  debuginfod_off,
> +  debuginfod_ask,
> +  NULL,
> +};
> +
> +static const char *debuginfod_enable = debuginfod_ask;
> +static std::string debuginfod_urls;
> +static bool debuginfod_debug = true;
> +
> +/* Show whether debuginfod is enabled.  */
> +
> +static void
> +show_debuginfod_command (struct ui_file *file, int from_tty,
> +			 struct cmd_list_element *cmd, const char *value)
> +{
> +   fprintf_unfiltered (file, _("Debuginfod functionality is currently set " \
> +		       "to \"%s\".\n"), debuginfod_enable);
> +}
> +
> +/* Set the URLs that debuginfod will query.  */
> +
> +static void
> +set_debuginfod_urls_command (const char *args, int from_tty,
> +			     struct cmd_list_element *c)
> +{
> +  if (setenv ("DEBUGINFOD_URLS", debuginfod_urls.c_str (), 1) != 0)
> +    warning (_("Unable to set debuginfod URLs: %s"), strerror (errno));
>
I believe we prefer to use safe_strerror (as used elsewhere in this file).

> +
> +/* Show the URLs that debuginfod will query.  */
> +
> +static void
> +show_debuginfod_urls_command (struct ui_file *file, int from_tty,
> +			      struct cmd_list_element *cmd, const char *value)
> +{
> +  if (debuginfod_urls.empty ())
> +    fprintf_unfiltered (file, _("You have not set debuginfod URLs.\n"));
> +  else
> +    fprintf_filtered (file, _("Debuginfod URLs are currently set to:\n%s\n"),
> +		      value);
> +}
> +
> +/* Show whether to display debuginfod debugging output.  */
> +
> +static void
> +show_debuginfod_debug_command (struct ui_file *file, int from_tty,
> +			       struct cmd_list_element *cmd, const char *value)
> +{
> +  fprintf_filtered (file, _("Debuginfod debugging output is %s.\n"), value);
> +}
> +
>   static int
>   progressfn (debuginfod_client *c, long cur, long total)
>   {
> @@ -195,3 +283,50 @@ debuginfod_debuginfo_query (const unsigned char *build_id,
>     return fd;
>   }
>   #endif
> +
> +/* Register debuginfod commands.  */
> +
> +void _initialize_debuginfod ();
> +void
> +_initialize_debuginfod ()
> +{
> +#ifdef HAVE_LIBDEBUGINFOD
> +  char *urls = getenv (DEBUGINFOD_URLS_ENV_VAR);
> +
> +  if (urls != nullptr)
> +    debuginfod_urls = urls;
> +
> +  add_setshow_enum_cmd ("debuginfod", class_run, debuginfod_cmd_names,
> +			&debuginfod_enable, _("\
> +Set whether debuginfod is enabled."), _("\
> +Show whether debuginfod is enabled."), _("\
> +Automatically query debuginfod servers for missing debuginfo, executables \
> +or source files. This command accepts the following arguments:\n\
> +  on  - debuginfod will be enabled\n\
> +  off - debuginfod will be disabled\n\
> +  ask - an interactive prompt will ask whether to enable debuginfod\n\
> +For interactive sessions, \"ask\" is the default. For non-interactive \
> +sessions, \"off\" is the default."),
> +			NULL, show_debuginfod_command, &setlist, &showlist);
> +
> +  add_setshow_string_noescape_cmd ("debuginfod-urls", class_run,
> +				   &debuginfod_urls, _("\
> +Set the list of debuginfod server URLs."), _("\
> +Show the list of debuginfod server URLs."), _("\
> +Manage the space-separated list of debuginfod server URLs that GDB will query \
> +when missing debuginfo, executables or source files.\nThe default value is \
> +copied from $DEBUGINFOD_URLS."),
> +				   set_debuginfod_urls_command,
> +				   show_debuginfod_urls_command,
> +				   &setlist, &showlist);
> +
> +  add_setshow_boolean_cmd ("debuginfod", class_maintenance,
> +			   &debuginfod_debug, _("\
> +Set debuginfod debugging."), _("\
> +Show debuginfod debugging."), _("\
> +When set, display debugging output for debuginfod. \
> +Displayed by default"),
> +			   NULL, show_debuginfod_debug_command,
> +			   &setdebuglist, &showdebuglist);
> +#endif
> +}

Just a question for others: Is this the normal style used in GDB? E.g., the
python command is always available:
(gdb) python print(1)
Python scripting is not supported in this copy of GDB.

Like command/command-line options, I'd like us to be vigilant of consistency.
[Maybe that's just my personal thing, though.]

It's just a question -- I am not asking for any changes.

With the safe_strerror change, though, I recommend a maintainer review
this patch for final approval(s).

Keith