public inbox for gdb-patches@sourceware.org
 help / color / mirror / Atom feed
From: Andrew Burgess <andrew.burgess@embecosm.com>
To: Tom de Vries <tdevries@suse.de>
Cc: Marco Barisione <mbarisione@undo.io>, Tom Tromey <tom@tromey.com>,
	GDB patches mailing list <gdb-patches@sourceware.org>
Subject: Re: [RFC][gdb/cli] Ignore error in gdb command script
Date: Tue, 18 May 2021 15:23:46 +0100	[thread overview]
Message-ID: <20210518142346.GS3067949@embecosm.com> (raw)
In-Reply-To: <eac69a98-953d-2458-8db1-9ebdafd4cbae@suse.de>

* Tom de Vries <tdevries@suse.de> [2021-05-18 15:57:42 +0200]:

> On 5/18/21 1:12 PM, Marco Barisione wrote:
> > On 18 May 2021, at 10:59, Tom de Vries <tdevries@suse.de> wrote:
> >> Hi,
> >>
> >> While trying to reproduce a failing test-case from the testsuite on the
> >> command line using a gdb command script, I ran into the problem that a command
> >> failed which stopped script execution.
> >>
> >> I could work around this by splitting the script at each error, but I realized
> >> it would be nice if I could tell gdb to ignore the error.
> >>
> >> Inspired by make, I chose the '-' prefix.
> > 
> > As MI commands are prefixed by “-“, isn’t there a risk of confusion?
> > 
> 
> Ah, right, I tend to forget about MI, good point.
> 
> > There’s also a “-” command (see tui/tui-win.c) which will stop working with
> > your patch.
> 
> I see, that's:
> ...
> $ gdb -q -batch -ex "help -"
> Scroll window backward.
> Usage: - [N] [WIN]
> Scroll window WIN N lines backwards.  Both WIN and N are optional, N
> defaults to 1, and WIN defaults to the currently focused window.
> ...
> 
> FWIW, did not find any documentation for this command.
> 
> Anyway, the '-' prefix seems a poor choice.
> 
> I've updated the patch to implement the ignore-errors idiom natively (so
> it also works without python).
> 
> Also, I've managed to fix the hang, by copying some code from
> execute_gdb_command.
> 
> Thanks,
> - Tom
> 

> [gdb/cli] Add ignore-errors command
> 
> While trying to reproduce a failing test-case from the testsuite on the
> command line using a gdb command script, I ran into the problem that a command
> failed which stopped script execution.
> 
> I could work around this by splitting the script at each error, but I realized
> it would be nice if I could tell gdb to ignore the error.
> 
> A python workaround ignore-errors exists, mentioned here (
> https://sourceware.org/legacy-ml/gdb/2010-06/msg00100.html ).  This also
> mentions a bugzilla entry that adds "exception handling to the gdb command
> language" but I was not able to find that.
> 
> This patch adds native ignore-errors support (so no python needed).
> 
> So with this script:
> ...
> $ cat script.gdb
> ignore-errors run
> echo here
> ...
> we have:
> ...
> $ gdb -q -batch -x script.gdb
> here$
> ...
> 
> We could also implement this as first-class command, like so:
> ...
> static void
> ignore_errors_command (const char *args, int from_tty)
> {
>   try
>     { execute_command (args, from_tty); }
>   catch (const gdb_exception_error &ex)
>     { async_enable_stdin (); }
> }
> ...
> and:
> ...
>   add_cmd ("ignore-errors", class_support, ignore_errors_command,
>            source_help_text, &cmdlist);
> ...
> but that means we go twice through execute_command.  This seems cleaner, at
> least in that aspect.

The benefit of implementing as a first class command is that we can
also get command completion on the second-level command.  See, for
example, the existing "with" command, so I can do:

  (gdb) with prin<TAB>

and get:

  (gdb) with print

It would be great it we got this for ignore-errors too.

Thanks,
Andrew



> 
> ---
>  gdb/event-top.c | 29 +++++++++++++++++++++++++++--
>  1 file changed, 27 insertions(+), 2 deletions(-)
> 
> diff --git a/gdb/event-top.c b/gdb/event-top.c
> index 002a7dc95e0..9d36b550dda 100644
> --- a/gdb/event-top.c
> +++ b/gdb/event-top.c
> @@ -583,13 +583,38 @@ command_handler (const char *command)
>    /* Do not execute commented lines.  */
>    for (c = command; *c == ' ' || *c == '\t'; c++)
>      ;
> -  if (c[0] != '#')
> +  if (c[0] == '#')
> +    return;
> +
> +  /* Detect and skip "ignore-errors". */
> +  const char * s = "ignore-errors";
> +  const size_t s_len = strlen (s);
> +  bool ignore_errors_p
> +    = (strncmp (c, s, s_len) == 0
> +       && strlen (c) > s_len
> +       && (c[s_len] == ' ' || c[s_len] == '\t'));
> +  if (ignore_errors_p)
>      {
> -      execute_command (command, ui->instream == ui->stdin_stream);
> +      c += s_len;
> +      for (; *c == ' ' || *c == '\t'; c++)
> +	;
> +    }
> +
> +  try
> +    {
> +      execute_command (c, ui->instream == ui->stdin_stream);
>  
>        /* Do any commands attached to breakpoint we stopped at.  */
>        bpstat_do_actions ();
>      }
> +  catch (const gdb_exception_error &ex)
> +    {
> +      if (!ignore_errors_p)
> +	throw;
> +
> +      /* See also execute_gdb_command.  */
> +      async_enable_stdin ();
> +    }
>  }
>  
>  /* Append RL, an input line returned by readline or one of its


      parent reply	other threads:[~2021-05-18 14:23 UTC|newest]

Thread overview: 15+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-05-18  9:59 Tom de Vries
2021-05-18 11:12 ` Marco Barisione
2021-05-18 13:57   ` Tom de Vries
2021-05-18 14:10     ` Simon Marchi
2021-05-18 14:42       ` Tom Tromey
2021-05-18 15:16         ` Tom de Vries
2021-05-18 15:32           ` Tom Tromey
2021-05-19  6:25             ` [PATCH][gdb/cli] Add ignore-errors command Tom de Vries
2021-05-19  7:32               ` Andreas Schwab
2021-05-19  7:36                 ` Tom de Vries
2021-05-18 19:16           ` [RFC][gdb/cli] Ignore error in gdb command script Philippe Waroquiers
2021-05-18 21:59             ` Tom de Vries
2021-05-20  8:02               ` Philippe Waroquiers
2021-05-20 15:01                 ` Simon Marchi
2021-05-18 14:23     ` Andrew Burgess [this message]

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=20210518142346.GS3067949@embecosm.com \
    --to=andrew.burgess@embecosm.com \
    --cc=gdb-patches@sourceware.org \
    --cc=mbarisione@undo.io \
    --cc=tdevries@suse.de \
    --cc=tom@tromey.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;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).