[PATCHv6,06/10] gdb: parse pending breakpoint thread/task immediately

Message ID 2b291071a7869e0ca69aa89d3f197c68512439c1.1701513409.git.aburgess@redhat.com
State New
Headers
Series [PATCHv6,01/10] gdb: create_breakpoint: add asserts and additional comments |

Checks

Context Check Description
linaro-tcwg-bot/tcwg_gdb_build--master-aarch64 success Testing passed
linaro-tcwg-bot/tcwg_gdb_build--master-arm success Testing passed
linaro-tcwg-bot/tcwg_gdb_check--master-aarch64 fail Testing failed
linaro-tcwg-bot/tcwg_gdb_check--master-arm success Testing passed

Commit Message

Andrew Burgess Dec. 2, 2023, 10:42 a.m. UTC
  Eli has already approved the docs changes in this commit.

---

The initial motivation for this commit was to allow thread or inferior
specific breakpoints to only be inserted within the appropriate
inferior's program-space.  The benefit of this is that inferiors for
which the breakpoint does not apply will no longer need to stop, and
then resume, for such breakpoints.  This commit does not make this
change, but is a refactor to allow this to happen in a later commit.

The problem we currently have is that when a thread-specific (or
inferior-specific) breakpoint is created, the thread (or inferior)
number is only parsed by calling find_condition_and_thread_for_sals.
This function is only called for non-pending breakpoints, and requires
that we know the locations at which the breakpoint will be placed (for
expression checking in case the breakpoint is also conditional).

A consequence of this is that by the time we figure out the breakpoint
is thread-specific we have already looked up locations in all program
spaces.  This feels wasteful -- if we knew the thread-id earlier then
we could reduce the work GDB does by only looking up locations within
the program space for which the breakpoint applies.

Another consequence of how find_condition_and_thread_for_sals is
called is that pending breakpoints don't currently know they are
thread-specific, nor even that they are conditional!  Additionally, by
delaying parsing the thread-id, pending breakpoints can be created for
non-existent threads, this is different to how non-pending
breakpoints are handled, so I can do this:

  $ gdb -q ./gdb/testsuite/outputs/gdb.multi/pending-bp/pending-bp
  Reading symbols from ./gdb/testsuite/outputs/gdb.multi/pending-bp/pending-bp...
  (gdb) break foo thread 99
  Function "foo" not defined.
  Make breakpoint pending on future shared library load? (y or [n]) y
  Breakpoint 1 (foo thread 99) pending.
  (gdb) r
  Starting program: /tmp/gdb/testsuite/outputs/gdb.multi/pending-bp/pending-bp
  [Thread debugging using libthread_db enabled]
  Using host libthread_db library "/lib64/libthread_db.so.1".
  Error in re-setting breakpoint 1: Unknown thread 99.
  [Inferior 1 (process 3329749) exited normally]
  (gdb)

GDB only checked the validity of 'thread 99' at the point the 'foo'
location became non-pending.  In contrast, if I try this:

  $ gdb -q ./gdb/testsuite/outputs/gdb.multi/pending-bp/pending-bp
  Reading symbols from ./gdb/testsuite/outputs/gdb.multi/pending-bp/pending-bp...
  (gdb) break main thread 99
  Unknown thread 99.
  (gdb)

GDB immediately checks if 'thread 99' exists.  I think inconsistencies
like this are confusing, and should be fixed if possible.

In this commit the create_breakpoint function is updated so that the
extra_string, which contains the thread, inferior, task, and/or
condition information, is parsed immediately, even for pending
breakpoints.

Obviously, the condition still can't be validated until the breakpoint
becomes non-pending, but the thread, inferior, and task information
can be pulled from the extra-string, and can be validated early on,
even for pending breakpoints.  The -force-condition flag is also
parsed as part of this early parsing change.

There are a couple of benefits to doing this:

1. Printing of breakpoints is more consistent now.  Consider creating
   a conditional breakpoint before this commit:

    (gdb) set breakpoint pending on
    (gdb) break pendingfunc if (0)
    Function "pendingfunc" not defined.
    Breakpoint 1 (pendingfunc if (0)) pending.
    (gdb) break main if (0)
    Breakpoint 2 at 0x401198: file /tmp/hello.c, line 18.
    (gdb) info breakpoints
    Num     Type           Disp Enb Address            What
    1       breakpoint     keep y   <PENDING>          pendingfunc if (0)
    2       breakpoint     keep y   0x0000000000401198 in main at /tmp/hello.c:18
            stop only if (0)
    (gdb)

   And after this commit:

    (gdb) set breakpoint pending on
    (gdb) break pendingfunc if (0)
    Function "pendingfunc" not defined.
    Breakpoint 1 (pendingfunc) pending.
    (gdb) break main if (0)
    Breakpoint 2 at 0x401198: file /home/andrew/tmp/hello.c, line 18.
    (gdb) info breakpoints
    Num     Type           Disp Enb Address            What
    1       breakpoint     keep y   <PENDING>          pendingfunc
            stop only if (0)
    2       breakpoint     keep y   0x0000000000401198 in main at /home/andrew/tmp/hello.c:18
            stop only if (0)
    (gdb)

   Notice that the display of the condition is now the same for the
   pending and non-pending breakpoints.

   The same is true for the thread, inferior, or task information in
   thread, inferior, or task specific breakpoints; this information is
   displayed on its own line rather than being part of the 'What'
   field.

2. We can check that the thread exists as soon as the pending
   breakpoint is created.  Currently there is a weird difference
   between pending and non-pending breakpoints when creating a
   thread-specific breakpoint.

   A pending thread-specific breakpoint only checks its thread when it
   becomes non-pending, at which point the thread the breakpoint was
   intended for might have exited.  Here's the behaviour before this
   commit:

    (gdb) set breakpoint pending on
    (gdb) break foo thread 2
    Function "foo" not defined.
    Breakpoint 2 (foo thread 2) pending.
    (gdb) c
    Continuing.
    [Thread 0x7ffff7c56700 (LWP 2948835) exited]
    Error in re-setting breakpoint 2: Unknown thread 2.
    [Inferior 1 (process 2948832) exited normally]
    (gdb)

   Notice the 'Error in re-setting breakpoint 2: Unknown thread 2.'
   line, this was triggered when GDB tried to make the breakpoint
   non-pending, and GDB discovers that the thread no longer exists.

   Compare that to the behaviour after this commit:

    (gdb) set breakpoint pending on
    (gdb) break foo thread 2
    Function "foo" not defined.
    Breakpoint 2 (foo) pending.
    (gdb) c
    Continuing.
    [Thread 0x7ffff7c56700 (LWP 2949243) exited]
    Thread-specific breakpoint 2 deleted - thread 2 no longer in the thread list.
    [Inferior 1 (process 2949240) exited normally]
    (gdb)

   Now the behaviour for pending breakpoints is identical to
   non-pending breakpoints, the thread specific breakpoint is removed
   as soon as the thread the breakpoint is associated with exits.

   There is an additional change; when the pending breakpoint is
   created prior to this patch we see this line:

     Breakpoint 2 (foo thread 2) pending.

   While after this patch we get this line:

     Breakpoint 2 (foo) pending.

   Notice that 'thread 2' has disappeared.  This might look like a
   regression, but I don't think it is.  That we said 'thread 2'
   before was just a consequence of the lazy parsing of the breakpoint
   specification, while with this patch GDB understands, and has
   parsed away the 'thread 2' bit of the spec.  If folk think the old
   information was useful then this would be trivial to add back in
   code_breakpoint::say_where.

As a result of this commit the breakpoints 'extra_string' field is now
only used by bp_dprintf type breakpoints to hold the printf format and
arguments.  This string should always be empty for other breakpoint
types.  This allows me to clean up some code in
print_breakpoint_location.

In code_breakpoint::code_breakpoint I'm able to change an error case
into an assert.  This is because this error is now handled earlier in
create_breakpoint.  As a result we know that by this point, the
extra_string will always be nullptr for anything other than a
bp_dprintf style breakpoint.

The find_condition_and_thread_for_sals function is now no longer
needed, this was previously doing the delayed splitting of the extra
string into thread, task, and condition, but this is now all done in
create_breakpoint, so find_condition_and_thread_for_sals can be
deleted, and the code that calls this in
code_breakpoint::location_spec_to_sals can be removed.  With this
update this code would only ever be reached for bp_dprintf style
breakpoints, and in these cases the extra_string should not contain
anything other than format and args.

The most interesting changes are all in create_breakpoint and in the
new file break-cond-parse.c.  We have a new block of code early on in
create_breakpoint that is responsible for splitting the extra_string
into its component parts by calling create_breakpoint_parse_arg_string
a function in the new break-cond-parse.c file.  This means that some
of the later code can be simplified a little.

The new break-cond-parse.c file implements the splitting up the
extra_string and finding all the parts, as well as some self-tests for
the new function.

Finally, now we know all the details, these can be stored within the
breakpoint object if we end up creating a deferred breakpoint.
Additionally, if we are creating a deferred bp_dprintf we can parse
the extra_string to build the printf command.

The implementation here aims to maintain backwards compatibility as
much as possible, this means that:

  1. We support abbreviations of 'thread', 'task', and 'inferior' in
  some places on the breakpoint line.  The handling of abbreviations
  has (before this patch) been a little weird, so this works:

  (gdb) break *main th 1

  And creates a breakpoint at '*main' for thread 1 only, while this
  does not work:

  (gdb) break main th 1

  In this case GDB will try to find the symbol 'main th 1'.  This
  weirdness exists before and after this patch.

  2. The handling of '-force-condition' is odd, if this flag appears
  immediately after a condition then it will be treated as part of the
  condition, e.g.:

  (gdb) break main if 0 -force-condition
  No symbol "force" in current context.

  But we are fine with these alternatives:

  (gdb) break main if 0 thread 1 -force-condition
  (gdb) break main -force-condition if 0

  Again, this is just a quirk of how the breakpoint line used to be
  parsed, but I've chosen to maintain this for backward
  compatibility.  Maybe in the future we might wish to consider
  changing this behaviour, but I'd rather do that as a separate commit
  later on, if that was of interest to people.

Reviewed-By: Eli Zaretskii <eliz@gnu.org>
---
 gdb/Makefile.in                               |   2 +
 gdb/NEWS                                      |   4 +
 gdb/break-cond-parse.c                        | 463 ++++++++++++++++++
 gdb/break-cond-parse.h                        |  49 ++
 gdb/breakpoint.c                              | 372 ++++----------
 gdb/testsuite/gdb.ada/tasks.exp               |   6 +-
 gdb/testsuite/gdb.base/condbreak.exp          |  57 ++-
 gdb/testsuite/gdb.base/pending.exp            |  23 +-
 gdb/testsuite/gdb.linespec/explicit.exp       |   4 +-
 gdb/testsuite/gdb.mi/mi-dprintf-pending.exp   |   3 +-
 .../gdb.multi/inferior-specific-bp.exp        |   4 +-
 .../gdb.threads/del-pending-thread-bp-lib.c   |  22 +
 .../gdb.threads/del-pending-thread-bp.c       |  85 ++++
 .../gdb.threads/del-pending-thread-bp.exp     | 108 ++++
 14 files changed, 900 insertions(+), 302 deletions(-)
 create mode 100644 gdb/break-cond-parse.c
 create mode 100644 gdb/break-cond-parse.h
 create mode 100644 gdb/testsuite/gdb.threads/del-pending-thread-bp-lib.c
 create mode 100644 gdb/testsuite/gdb.threads/del-pending-thread-bp.c
 create mode 100644 gdb/testsuite/gdb.threads/del-pending-thread-bp.exp
  

Comments

Tankut Baris Aktemur Dec. 5, 2023, 3:09 p.m. UTC | #1
On Saturday, December 2, 2023 11:42 AM, Andrew Burgess wrote:
> diff --git a/gdb/break-cond-parse.c b/gdb/break-cond-parse.c
> new file mode 100644
> index 00000000000..b59bd7aeeec
> --- /dev/null
> +++ b/gdb/break-cond-parse.c
> @@ -0,0 +1,463 @@
> +/* Copyright (C) 2023 Free Software Foundation, Inc.
> +
> +   This file is part of GDB.
> +
> +   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/>.  */
> +
> +#include "defs.h"
> +#include "gdbsupport/gdb_assert.h"
> +#include "gdbsupport/selftest.h"
> +#include "test-target.h"
> +#include "scoped-mock-context.h"
> +#include "break-cond-parse.h"
> +#include "tid-parse.h"
> +#include "ada-lang.h"
> +
> +/* When parsing tokens from a string, which direction are we parsing?
> +
> +   Given the following string and pointer 'ptr':
> +
> +	ABC DEF GHI JKL
> +	       ^
> +	       ptr
> +
> +   Parsing 'forward' will return the token 'GHI' and update 'ptr' to point
> +   between GHI and JKL.  Parsing 'backward' will return the token 'DEF' and
> +   update 'ptr' to point between ABC and DEF.
> +*/
> +
> +enum class parse_direction
> +{
> +  /* Parse the next token forwards.  */
> +  forward,
> +
> +  /* Parse the previous token backwards.  */
> +  backward
> +};
> +
> +struct token
> +{
> +  /* Create a new token.  START points to the first character of the new
> +     token, while END points at the last character of the new token.
> +
> +     Neither START or END can be nullptr, and both START and END must point
> +     into the same C style string (i.e. there should be no null character
> +     between START and END).  END must be equal to, or greater than START,
> +     that is, it is not possible to create a zero length token.  */
> +
> +  token (const char *start, const char *end)

Have you considered using std::string_view?

> +    : m_start (start),
> +      m_end (end)
> +  {
> +    gdb_assert (m_end >= m_start);
> +    gdb_assert (m_start + strlen (m_start) > m_end);
> +  }
> +
> +  /* Return a pointer to the first character of this token.  */
> +  const char *start () const
> +  {
> +    return m_start;
> +  }
> +
> +  /* Return a pointer to the last character of this token.  */
> +  const char *end () const
> +  {
> +    return m_end;
> +  }
> +
> +  /* Return the length of the token.  */
> +  size_t length () const
> +  {
> +    /* The + 1 is because the character at m_end is part of the token.  */
> +    return m_end - m_start + 1;
> +  }
> +
> +  /* Return true if this token matches STR.  */
> +  bool matches (const char *str) const
> +  {
> +    return strncmp (m_start, str, length ()) == 0;
> +  }
> +
> +private:
> +  /* The first character of this token.  */
> +  const char *m_start;
> +
> +  /* The last character of this token.  */
> +  const char *m_end;
> +};
> +
> +/* Find the next token in DIRECTION from *CURR.  */
> +
> +static token
> +find_next_token (const char **curr, parse_direction direction)
> +{
> +  const char *tok_start, *tok_end;
> +
> +  gdb_assert (**curr != '\0');
> +
> +  if (direction == parse_direction::forward)
> +    {
> +      *curr = skip_spaces (*curr);
> +      tok_start = *curr;
> +      *curr = skip_to_space (*curr);
> +      tok_end = *curr - 1;
> +    }
> +  else
> +    {
> +      gdb_assert (direction == parse_direction::backward);
> +
> +      while (isspace (**curr))
> +	--(*curr);
> +
> +      tok_end = *curr;
> +
> +      while (!isspace (**curr))
> +	--(*curr);
> +
> +      tok_start = (*curr) + 1;
> +    }
> +
> +  return token (tok_start, tok_end);
> +}
> +
> +/* See break-cond-parse.h.  */
> +
> +void
> +create_breakpoint_parse_arg_string
> +  (const char *tok, gdb::unique_xmalloc_ptr<char> *cond_string,
> +   int *thread, int *inferior, int *task,
> +   gdb::unique_xmalloc_ptr<char> *rest, bool *force)
> +{
> +  /* Set up the defaults.  */
> +  cond_string->reset ();
> +  rest->reset ();
> +  *thread = -1;
> +  *inferior = -1;
> +  *task = -1;
> +  *force = false;
> +
> +  if (tok == nullptr)
> +    return;
> +
> +  /* The "-force-condition" flag gets some special treatment.  If this
> +     token is immediately after the condition string then we treat this as
> +     part of the condition string rather than a separate token.  This is
> +     just a quirk of how this token used to be parsed, and has been
> +     retained for backwards compatibility.  Maybe this should be updated
> +     in the future.  */

(I was the author of the -force-condition patch.)
The quirks about the -force-condition flag were not a deliberate feature.
For whatever it's worth, I think it's OK to de-prioritize backwards
compatibility for this flag.  I doubt that it would impact many users.

In fact, in a post-merge comment, Pedro and Tom had suggested converting
the flag to an option rather than a keyword:

https://sourceware.org/pipermail/gdb-patches/2020-October/172952.html
https://sourceware.org/pipermail/gdb-patches/2020-December/173802.html
https://sourceware.org/pipermail/gdb-patches/2020-December/173999.html

I (embarrassingly) did not have the opportunity to get back to it.  Maybe
it's now the correct time to address the concerns?  I can gladly help.
What do you think?

...
>    try
>      {
>        ops->create_sals_from_location_spec (locspec, &canonical);
> @@ -9285,6 +9122,13 @@ create_breakpoint (struct gdbarch *gdbarch,
>  	throw;
>      }
> 
> +  /* The only bp_dprintf should have anything in EXTRA_STRING_COPY by this

This sentence sounds odd.  Did you mean "Only the bp_dprintf type should have..."?

> +     point.  For all other breakpoints this indicates an error.  We could
> +     place this check earlier in the function, but we prefer to see errors
> +     from the location spec parser before we see this error message.  */
> +  if (type_wanted != bp_dprintf && extra_string_copy.get () != nullptr)
> +    error (_("Garbage '%s' at end of command"), extra_string_copy.get ());
> +
>    if (!pending && canonical.lsals.empty ())
>      return 0;
> 

Regards
-Baris


Intel Deutschland GmbH
Registered Address: Am Campeon 10, 85579 Neubiberg, Germany
Tel: +49 89 99 8853-0, www.intel.de <http://www.intel.de>
Managing Directors: Christin Eisenschmid, Sharon Heck, Tiffany Doon Silva  
Chairperson of the Supervisory Board: Nicole Lau
Registered Office: Munich
Commercial Register: Amtsgericht Muenchen HRB 186928
  
Andrew Burgess Dec. 13, 2023, 1:51 p.m. UTC | #2
"Aktemur, Tankut Baris" <tankut.baris.aktemur@intel.com> writes:

> On Saturday, December 2, 2023 11:42 AM, Andrew Burgess wrote:
>> diff --git a/gdb/break-cond-parse.c b/gdb/break-cond-parse.c
>> new file mode 100644
>> index 00000000000..b59bd7aeeec
>> --- /dev/null
>> +++ b/gdb/break-cond-parse.c
>> @@ -0,0 +1,463 @@
>> +/* Copyright (C) 2023 Free Software Foundation, Inc.
>> +
>> +   This file is part of GDB.
>> +
>> +   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/>.  */
>> +
>> +#include "defs.h"
>> +#include "gdbsupport/gdb_assert.h"
>> +#include "gdbsupport/selftest.h"
>> +#include "test-target.h"
>> +#include "scoped-mock-context.h"
>> +#include "break-cond-parse.h"
>> +#include "tid-parse.h"
>> +#include "ada-lang.h"
>> +
>> +/* When parsing tokens from a string, which direction are we parsing?
>> +
>> +   Given the following string and pointer 'ptr':
>> +
>> +	ABC DEF GHI JKL
>> +	       ^
>> +	       ptr
>> +
>> +   Parsing 'forward' will return the token 'GHI' and update 'ptr' to point
>> +   between GHI and JKL.  Parsing 'backward' will return the token 'DEF' and
>> +   update 'ptr' to point between ABC and DEF.
>> +*/
>> +
>> +enum class parse_direction
>> +{
>> +  /* Parse the next token forwards.  */
>> +  forward,
>> +
>> +  /* Parse the previous token backwards.  */
>> +  backward
>> +};
>> +
>> +struct token
>> +{
>> +  /* Create a new token.  START points to the first character of the new
>> +     token, while END points at the last character of the new token.
>> +
>> +     Neither START or END can be nullptr, and both START and END must point
>> +     into the same C style string (i.e. there should be no null character
>> +     between START and END).  END must be equal to, or greater than START,
>> +     that is, it is not possible to create a zero length token.  */
>> +
>> +  token (const char *start, const char *end)
>
> Have you considered using std::string_view?
>
>> +    : m_start (start),
>> +      m_end (end)
>> +  {
>> +    gdb_assert (m_end >= m_start);
>> +    gdb_assert (m_start + strlen (m_start) > m_end);
>> +  }
>> +
>> +  /* Return a pointer to the first character of this token.  */
>> +  const char *start () const
>> +  {
>> +    return m_start;
>> +  }
>> +
>> +  /* Return a pointer to the last character of this token.  */
>> +  const char *end () const
>> +  {
>> +    return m_end;
>> +  }
>> +
>> +  /* Return the length of the token.  */
>> +  size_t length () const
>> +  {
>> +    /* The + 1 is because the character at m_end is part of the token.  */
>> +    return m_end - m_start + 1;
>> +  }
>> +
>> +  /* Return true if this token matches STR.  */
>> +  bool matches (const char *str) const
>> +  {
>> +    return strncmp (m_start, str, length ()) == 0;
>> +  }
>> +
>> +private:
>> +  /* The first character of this token.  */
>> +  const char *m_start;
>> +
>> +  /* The last character of this token.  */
>> +  const char *m_end;
>> +};
>> +
>> +/* Find the next token in DIRECTION from *CURR.  */
>> +
>> +static token
>> +find_next_token (const char **curr, parse_direction direction)
>> +{
>> +  const char *tok_start, *tok_end;
>> +
>> +  gdb_assert (**curr != '\0');
>> +
>> +  if (direction == parse_direction::forward)
>> +    {
>> +      *curr = skip_spaces (*curr);
>> +      tok_start = *curr;
>> +      *curr = skip_to_space (*curr);
>> +      tok_end = *curr - 1;
>> +    }
>> +  else
>> +    {
>> +      gdb_assert (direction == parse_direction::backward);
>> +
>> +      while (isspace (**curr))
>> +	--(*curr);
>> +
>> +      tok_end = *curr;
>> +
>> +      while (!isspace (**curr))
>> +	--(*curr);
>> +
>> +      tok_start = (*curr) + 1;
>> +    }
>> +
>> +  return token (tok_start, tok_end);
>> +}
>> +
>> +/* See break-cond-parse.h.  */
>> +
>> +void
>> +create_breakpoint_parse_arg_string
>> +  (const char *tok, gdb::unique_xmalloc_ptr<char> *cond_string,
>> +   int *thread, int *inferior, int *task,
>> +   gdb::unique_xmalloc_ptr<char> *rest, bool *force)
>> +{
>> +  /* Set up the defaults.  */
>> +  cond_string->reset ();
>> +  rest->reset ();
>> +  *thread = -1;
>> +  *inferior = -1;
>> +  *task = -1;
>> +  *force = false;
>> +
>> +  if (tok == nullptr)
>> +    return;
>> +
>> +  /* The "-force-condition" flag gets some special treatment.  If this
>> +     token is immediately after the condition string then we treat this as
>> +     part of the condition string rather than a separate token.  This is
>> +     just a quirk of how this token used to be parsed, and has been
>> +     retained for backwards compatibility.  Maybe this should be updated
>> +     in the future.  */
>
> (I was the author of the -force-condition patch.)
> The quirks about the -force-condition flag were not a deliberate feature.
> For whatever it's worth, I think it's OK to de-prioritize backwards
> compatibility for this flag.  I doubt that it would impact many users.
>
> In fact, in a post-merge comment, Pedro and Tom had suggested converting
> the flag to an option rather than a keyword:
>
> https://sourceware.org/pipermail/gdb-patches/2020-October/172952.html
> https://sourceware.org/pipermail/gdb-patches/2020-December/173802.html
> https://sourceware.org/pipermail/gdb-patches/2020-December/173999.html
>
> I (embarrassingly) did not have the opportunity to get back to it.  Maybe
> it's now the correct time to address the concerns?  I can gladly help.
> What do you think?

I think this sounds like a good idea, but ....

... I took a quick look at how this might be done, and the problem I see
is that the current 'break' option parsing is buried within
string_to_explicit_location_spec, which is called from
string_to_location_spec, which is called from multiple commands.

Which makes sense, currently, all of the options for 'break' are about
specifying the location.

In contrast, -force-condition has nothing to do with the location, but
relates (as you know) to the condition.

We could handle -force-condition in string_to_explicit_location_spec,
but then commands like 'edit', which take a location, but not a
condition would accept -force-condition, and we'd have to have some
mechanism to the 'force' flag true/false value back out of
string_to_explicit_location_spec, either passing a container around, or
storing the 'force' state within the locspec... this all sounds like the
wrong approach.

Better I think would be to pull the option parsing out of
string_to_explicit_location_spec, and move it into the individual
commands.  What we'd ideally want is to convert the current bespoke
option parsing (for location spec options) to use GDB's generic option
parsing mechanism.  It's possible to bind multiple
gdb::option::option_def into a single gdb::option::option_def_group, so
we can imagine that the 'break' command would use a common
gdb::option::option_def from location.h, which defines all the location
spec arguments, and then a separate gdb::option::option_def which adds
the -force-condition flag.  The only issue then is that we'd need to
pass the location spec related arguments into
string_to_explicit_location_spec somehow....

Phew.  Now the next problem is that the gdb::option::option_def contains
a list of gdb::option::*_option_def objects, the precise type of which
defines how the options are parsed.  It we look at how the arguments to
things like -function are parsed, these get special language specific
handling, which the current option mechanism doesn't support, so I think
we'd need to add that, which will mean at least extending the option
process, or maybe even some reworking of the option handling...

All that is to say that I agree with Pedro that this would be better
done as an real option ... but I really don't want to tie this work to
that refactoring if at all possible...

Thanks,
Andrew

>
> ...
>>    try
>>      {
>>        ops->create_sals_from_location_spec (locspec, &canonical);
>> @@ -9285,6 +9122,13 @@ create_breakpoint (struct gdbarch *gdbarch,
>>  	throw;
>>      }
>> 
>> +  /* The only bp_dprintf should have anything in EXTRA_STRING_COPY by this
>
> This sentence sounds odd.  Did you mean "Only the bp_dprintf type should have..."?
>
>> +     point.  For all other breakpoints this indicates an error.  We could
>> +     place this check earlier in the function, but we prefer to see errors
>> +     from the location spec parser before we see this error message.  */
>> +  if (type_wanted != bp_dprintf && extra_string_copy.get () != nullptr)
>> +    error (_("Garbage '%s' at end of command"), extra_string_copy.get ());
>> +
>>    if (!pending && canonical.lsals.empty ())
>>      return 0;
>> 
>
> Regards
> -Baris
>
>
> Intel Deutschland GmbH
> Registered Address: Am Campeon 10, 85579 Neubiberg, Germany
> Tel: +49 89 99 8853-0, www.intel.de <http://www.intel.de>
> Managing Directors: Christin Eisenschmid, Sharon Heck, Tiffany Doon Silva  
> Chairperson of the Supervisory Board: Nicole Lau
> Registered Office: Munich
> Commercial Register: Amtsgericht Muenchen HRB 186928
  
Tankut Baris Aktemur Dec. 27, 2023, 12:23 p.m. UTC | #3
On Wednesday, December 13, 2023 2:51 PM, Andrew Burgess wrote:
> >> +  /* The "-force-condition" flag gets some special treatment.  If this
> >> +     token is immediately after the condition string then we treat this as
> >> +     part of the condition string rather than a separate token.  This is
> >> +     just a quirk of how this token used to be parsed, and has been
> >> +     retained for backwards compatibility.  Maybe this should be updated
> >> +     in the future.  */
> >
> > (I was the author of the -force-condition patch.)
> > The quirks about the -force-condition flag were not a deliberate feature.
> > For whatever it's worth, I think it's OK to de-prioritize backwards
> > compatibility for this flag.  I doubt that it would impact many users.
> >
> > In fact, in a post-merge comment, Pedro and Tom had suggested converting
> > the flag to an option rather than a keyword:
> >
> > https://sourceware.org/pipermail/gdb-patches/2020-October/172952.html
> > https://sourceware.org/pipermail/gdb-patches/2020-December/173802.html
> > https://sourceware.org/pipermail/gdb-patches/2020-December/173999.html
> >
> > I (embarrassingly) did not have the opportunity to get back to it.  Maybe
> > it's now the correct time to address the concerns?  I can gladly help.
> > What do you think?
> 
> I think this sounds like a good idea, but ....
> 
> ... I took a quick look at how this might be done,

Thanks for taking a look!

> and the problem I see
> is that the current 'break' option parsing is buried within
> string_to_explicit_location_spec, which is called from
> string_to_location_spec, which is called from multiple commands.
> 
> Which makes sense, currently, all of the options for 'break' are about
> specifying the location.
> 
> In contrast, -force-condition has nothing to do with the location, but
> relates (as you know) to the condition.
> 
> We could handle -force-condition in string_to_explicit_location_spec,
> but then commands like 'edit', which take a location, but not a
> condition would accept -force-condition, and we'd have to have some
> mechanism to the 'force' flag true/false value back out of
> string_to_explicit_location_spec, either passing a container around, or
> storing the 'force' state within the locspec... this all sounds like the
> wrong approach.
> 
> Better I think would be to pull the option parsing out of
> string_to_explicit_location_spec, and move it into the individual
> commands.  What we'd ideally want is to convert the current bespoke
> option parsing (for location spec options) to use GDB's generic option
> parsing mechanism.  It's possible to bind multiple
> gdb::option::option_def into a single gdb::option::option_def_group, so
> we can imagine that the 'break' command would use a common
> gdb::option::option_def from location.h, which defines all the location
> spec arguments, and then a separate gdb::option::option_def which adds
> the -force-condition flag.  The only issue then is that we'd need to
> pass the location spec related arguments into
> string_to_explicit_location_spec somehow....
> 
> Phew.  Now the next problem is that the gdb::option::option_def contains
> a list of gdb::option::*_option_def objects, the precise type of which
> defines how the options are parsed.  It we look at how the arguments to
> things like -function are parsed, these get special language specific
> handling, which the current option mechanism doesn't support, so I think
> we'd need to add that, which will mean at least extending the option
> process, or maybe even some reworking of the option handling...
> 
> All that is to say that I agree with Pedro that this would be better
> done as an real option ... but I really don't want to tie this work to
> that refactoring if at all possible...

Yes, sure, given the non-trivial nature of that refactoring, this makes sense.

Thanks
-Baris


Intel Deutschland GmbH
Registered Address: Am Campeon 10, 85579 Neubiberg, Germany
Tel: +49 89 99 8853-0, www.intel.de <http://www.intel.de>
Managing Directors: Christin Eisenschmid, Sharon Heck, Tiffany Doon Silva  
Chairperson of the Supervisory Board: Nicole Lau
Registered Office: Munich
Commercial Register: Amtsgericht Muenchen HRB 186928
  

Patch

diff --git a/gdb/Makefile.in b/gdb/Makefile.in
index 3510577f090..425b41f5973 100644
--- a/gdb/Makefile.in
+++ b/gdb/Makefile.in
@@ -1025,6 +1025,7 @@  COMMON_SFILES = \
 	break-catch-sig.c \
 	break-catch-syscall.c \
 	break-catch-throw.c \
+	break-cond-parse.c \
 	breakpoint.c \
 	bt-utils.c \
 	btrace.c \
@@ -1293,6 +1294,7 @@  HFILES_NO_SRCDIR = \
 	bfd-target.h \
 	bfin-tdep.h \
 	block.h \
+	break-cond-parse.h \
 	breakpoint.h \
 	bsd-kvm.h \
 	bsd-uthread.h \
diff --git a/gdb/NEWS b/gdb/NEWS
index 1073e38dfc6..a3cbd34f1d0 100644
--- a/gdb/NEWS
+++ b/gdb/NEWS
@@ -9,6 +9,10 @@ 
 * GDB index now contains information about the main function.  This speeds up
   startup when it is being used for some large binaries.
 
+* For breakpoints that are created in the 'pending' state, any
+  'thread' or 'task' keywords are parsed at the time the breakpoint is
+  created, rather than at the time the breakpoint becomes non-pending.
+
 * Changed commands
 
 disassemble
diff --git a/gdb/break-cond-parse.c b/gdb/break-cond-parse.c
new file mode 100644
index 00000000000..b59bd7aeeec
--- /dev/null
+++ b/gdb/break-cond-parse.c
@@ -0,0 +1,463 @@ 
+/* Copyright (C) 2023 Free Software Foundation, Inc.
+
+   This file is part of GDB.
+
+   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/>.  */
+
+#include "defs.h"
+#include "gdbsupport/gdb_assert.h"
+#include "gdbsupport/selftest.h"
+#include "test-target.h"
+#include "scoped-mock-context.h"
+#include "break-cond-parse.h"
+#include "tid-parse.h"
+#include "ada-lang.h"
+
+/* When parsing tokens from a string, which direction are we parsing?
+
+   Given the following string and pointer 'ptr':
+
+	ABC DEF GHI JKL
+	       ^
+	       ptr
+
+   Parsing 'forward' will return the token 'GHI' and update 'ptr' to point
+   between GHI and JKL.  Parsing 'backward' will return the token 'DEF' and
+   update 'ptr' to point between ABC and DEF.
+*/
+
+enum class parse_direction
+{
+  /* Parse the next token forwards.  */
+  forward,
+
+  /* Parse the previous token backwards.  */
+  backward
+};
+
+struct token
+{
+  /* Create a new token.  START points to the first character of the new
+     token, while END points at the last character of the new token.
+
+     Neither START or END can be nullptr, and both START and END must point
+     into the same C style string (i.e. there should be no null character
+     between START and END).  END must be equal to, or greater than START,
+     that is, it is not possible to create a zero length token.  */
+
+  token (const char *start, const char *end)
+    : m_start (start),
+      m_end (end)
+  {
+    gdb_assert (m_end >= m_start);
+    gdb_assert (m_start + strlen (m_start) > m_end);
+  }
+
+  /* Return a pointer to the first character of this token.  */
+  const char *start () const
+  {
+    return m_start;
+  }
+
+  /* Return a pointer to the last character of this token.  */
+  const char *end () const
+  {
+    return m_end;
+  }
+
+  /* Return the length of the token.  */
+  size_t length () const
+  {
+    /* The + 1 is because the character at m_end is part of the token.  */
+    return m_end - m_start + 1;
+  }
+
+  /* Return true if this token matches STR.  */
+  bool matches (const char *str) const
+  {
+    return strncmp (m_start, str, length ()) == 0;
+  }
+
+private:
+  /* The first character of this token.  */
+  const char *m_start;
+
+  /* The last character of this token.  */
+  const char *m_end;
+};
+
+/* Find the next token in DIRECTION from *CURR.  */
+
+static token
+find_next_token (const char **curr, parse_direction direction)
+{
+  const char *tok_start, *tok_end;
+
+  gdb_assert (**curr != '\0');
+
+  if (direction == parse_direction::forward)
+    {
+      *curr = skip_spaces (*curr);
+      tok_start = *curr;
+      *curr = skip_to_space (*curr);
+      tok_end = *curr - 1;
+    }
+  else
+    {
+      gdb_assert (direction == parse_direction::backward);
+
+      while (isspace (**curr))
+	--(*curr);
+
+      tok_end = *curr;
+
+      while (!isspace (**curr))
+	--(*curr);
+
+      tok_start = (*curr) + 1;
+    }
+
+  return token (tok_start, tok_end);
+}
+
+/* See break-cond-parse.h.  */
+
+void
+create_breakpoint_parse_arg_string
+  (const char *tok, gdb::unique_xmalloc_ptr<char> *cond_string,
+   int *thread, int *inferior, int *task,
+   gdb::unique_xmalloc_ptr<char> *rest, bool *force)
+{
+  /* Set up the defaults.  */
+  cond_string->reset ();
+  rest->reset ();
+  *thread = -1;
+  *inferior = -1;
+  *task = -1;
+  *force = false;
+
+  if (tok == nullptr)
+    return;
+
+  /* The "-force-condition" flag gets some special treatment.  If this
+     token is immediately after the condition string then we treat this as
+     part of the condition string rather than a separate token.  This is
+     just a quirk of how this token used to be parsed, and has been
+     retained for backwards compatibility.  Maybe this should be updated
+     in the future.  */
+  std::optional<token> force_condition_token;
+
+  const char *cond_start = nullptr;
+  const char *cond_end = nullptr;
+  parse_direction direction = parse_direction::forward;
+  while (*tok != '\0')
+    {
+      /* Find the next token.  If moving backward and this token starts at
+	 the same location as the condition then we must have found the
+	 other end of the condition string -- we're done.  */
+      token t = find_next_token (&tok, direction);
+      if (direction == parse_direction::backward && t.start () <= cond_start)
+	{
+	  cond_end = t.end ();
+	  break;
+	}
+
+      /* We only have a single flag option to check for.  All the other
+	 options take a value so require an additional token to be found.
+	 Additionally, we require that this flag be at least '-f', we
+	 don't allow it to be abbreviated to '-'.  */
+      if (t.length () > 1 && t.matches ("-force-condition"))
+	{
+	  force_condition_token.emplace (t);
+	  continue;
+	}
+
+      /* Maybe the first token was the last token in the string.  If this
+	 is the case then we definitely can't try to extract a value
+	 token.  This also means that the token T is meaningless.  Reset
+	 TOK to point at the start of the unknown content and break out of
+	 the loop.  We'll record the unknown part of the string outside of
+	 the scanning loop (below).  */
+      if (direction == parse_direction::forward && *tok == '\0')
+	{
+	  tok = t.start ();
+	  break;
+	}
+
+      /* As before, find the next token and, if we are scanning backwards,
+	 check that we have not reached the start of the condition string.  */
+      token v = find_next_token (&tok, direction);
+      if (direction == parse_direction::backward && v.start () <= cond_start)
+	{
+	  /* Use token T here as that must also be part of the condition
+	     string.  */
+	  cond_end = t.end ();
+	  break;
+	}
+
+      /* When moving backward we will first parse the value token then the
+	 keyword token, so swap them now.  */
+      if (direction == parse_direction::backward)
+	std::swap (t, v);
+
+      /* Check for valid option in token T.  If we find a valid option then
+	 parse the value from the token V.  Except for 'if', that's handled
+	 differently.
+
+	 For the 'if' token we need to capture the entire condition
+	 string, so record the start of the condition string and then
+	 start scanning backwards looking for the end of the condition
+	 string.
+
+	 The order of these checks is important, at least the check for
+	 'thread' must occur before the check for 'task'.  We accept
+	 abbreviations of these token names, and 't' should resolve to
+	 'thread', which will only happen if we check 'thread' first.  */
+      if (direction == parse_direction::forward && t.matches ("if"))
+	{
+	  cond_start = v.start ();
+	  tok = tok + strlen (tok);
+	  gdb_assert (*tok == '\0');
+	  --tok;
+	  direction = parse_direction::backward;
+	  continue;
+	}
+      else if (t.matches ("thread"))
+	{
+	  if (*thread != -1)
+	    error(_("You can specify only one thread."));
+
+	  if (*task != -1 || *inferior != -1)
+	    error (_("You can specify only one of thread, inferior, or task."));
+
+	  const char *tmptok;
+	  thread_info *thr = parse_thread_id (v.start (), &tmptok);
+	  const char *expected_end = skip_spaces (v.end () + 1);
+	  if (tmptok != expected_end)
+	    error (_("Junk after thread keyword."));
+	  *thread = thr->global_num;
+	}
+      else if (t.matches ("inferior"))
+	{
+	  if (*inferior != -1)
+	    error(_("You can specify only one inferior."));
+
+	  if (*task != -1 || *thread != -1)
+	    error (_("You can specify only one of thread, inferior, or task."));
+
+	  char *tmptok;
+	  *inferior = strtol (v.start (), &tmptok, 0);
+	  const char *expected_end = v.end () + 1;
+	  if (tmptok != expected_end)
+	    error (_("Junk after inferior keyword."));
+	}
+      else if (t.matches ("task"))
+	{
+	  if (*task != -1)
+	    error(_("You can specify only one task."));
+
+	  if (*thread != -1 || *inferior != -1)
+	    error (_("You can specify only one of thread, inferior, or task."));
+
+	  char *tmptok;
+	  *task = strtol (v.start (), &tmptok, 0);
+	  const char *expected_end = v.end () + 1;
+	  if (tmptok != expected_end)
+	    error (_("Junk after task keyword."));
+	  if (!valid_task_id (*task))
+	    error (_("Unknown task %d."), *task);
+	}
+      else
+	{
+	  /* An unknown token.  If we are scanning forward then reset TOK
+	     to point at the start of the unknown content, we record this
+	     outside of the scanning loop (below).
+
+	     If we are scanning backward then unknown content is assumed to
+	     be the other end of the condition string, obviously, this is
+	     just a heuristic, we could be looking at a mistyped command
+	     line, but this will be spotted when the condition is
+	     eventually evaluated.
+
+	     Either way, no more scanning is required after this.  */
+	  if (direction == parse_direction::forward)
+	    tok = t.start ();
+	  else
+	    {
+	      gdb_assert (direction == parse_direction::backward);
+	      cond_end = v.end ();
+	    }
+	  break;
+	}
+    }
+
+  if (cond_start != nullptr)
+    {
+      /* If we found the start of a condition string then we should have
+	 switched to backward scan mode, and found the end of the condition
+	 string.  Capture the whole condition string into COND_STRING
+	 now.  */
+      gdb_assert (direction == parse_direction::backward);
+      gdb_assert (cond_end != nullptr);
+
+      /* If the first token after the condition string is the
+	 "-force-condition" token, then we merge the "-force-condition"
+	 token with the condition string and forget ever seeing the
+	 "-force-condition".  */
+      if (force_condition_token.has_value ())
+	{
+	  const char *next_token_start = skip_spaces (cond_end + 1);
+
+	  if (next_token_start == force_condition_token->start ())
+	    {
+	      cond_end = force_condition_token->end ();
+	      force_condition_token.reset ();
+	    }
+	}
+
+      /* The '+ 1' here is because COND_END points to the last character of
+	 the condition string rather than the null-character at the end of
+	 the condition string, and we need the string length here.  */
+      cond_string->reset (savestring (cond_start, cond_end - cond_start + 1));
+    }
+  else if (*tok != '\0')
+    {
+      /* If we didn't have a condition start pointer then we should still
+	 be in forward scanning mode.  If we didn't reach the end of the
+	 input string (TOK is not at the null character) then the rest of
+	 the input string is garbage that we didn't understand.
+
+	 Record the unknown content into REST.  The caller of this function
+	 will report this as an error later on.  We could report the error
+	 here, but we prefer to allow the caller to run other checks, and
+	 prioritise other errors before reporting this problem.  */
+      gdb_assert (direction == parse_direction::forward);
+      gdb_assert (cond_end == nullptr);
+      rest->reset (savestring (tok, strlen (tok)));
+    }
+
+  /* If we saw "-force-condition" then set the *FORCE flag.  Depending on
+     which path we took above we might have chosen to forget having seen
+     the "-force-condition" token.  */
+  if (force_condition_token.has_value ())
+    *force = true;
+}
+
+#if GDB_SELF_TEST
+
+namespace selftests {
+
+/* Run a single test of the create_breakpoint_parse_arg_string function.
+   INPUT is passed to create_breakpoint_parse_arg_string while all other
+   arguments are the expected output from
+   create_breakpoint_parse_arg_string.  */
+
+static void
+test (const char *input, const char *condition, int thread = -1,
+      int inferior = -1, int task = -1, bool force = false,
+      const char *rest = nullptr)
+{
+  gdb::unique_xmalloc_ptr<char> extracted_condition;
+  gdb::unique_xmalloc_ptr<char> extracted_rest;
+  int extracted_thread, extracted_inferior, extracted_task;
+  bool extracted_force_condition;
+  std::string exception_msg;
+
+  try
+    {
+      create_breakpoint_parse_arg_string (input, &extracted_condition,
+					  &extracted_thread,
+					  &extracted_inferior,
+					  &extracted_task, &extracted_rest,
+					  &extracted_force_condition);
+    }
+  catch (const gdb_exception_error &ex)
+    {
+      string_file buf;
+
+      exception_print (&buf, ex);
+      exception_msg = buf.release ();
+    }
+
+  if ((condition == nullptr) != (extracted_condition.get () == nullptr)
+      || (condition != nullptr
+	  && strcmp (condition, extracted_condition.get ()) != 0)
+      || (rest == nullptr) != (extracted_rest.get () == nullptr)
+      || (rest != nullptr && strcmp (rest, extracted_rest.get ()) != 0)
+      || thread != extracted_thread
+      || inferior != extracted_inferior
+      || task != extracted_task
+      || force != extracted_force_condition
+      || !exception_msg.empty ())
+    {
+      if (run_verbose ())
+	{
+	  debug_printf ("input: '%s'\n", input);
+	  debug_printf ("condition: '%s'\n", extracted_condition.get ());
+	  debug_printf ("rest: '%s'\n", extracted_rest.get ());
+	  debug_printf ("thread: %d\n", extracted_thread);
+	  debug_printf ("inferior: %d\n", extracted_inferior);
+	  debug_printf ("task: %d\n", extracted_task);
+	  debug_printf ("forced: %s\n",
+			extracted_force_condition ? "true" : "false");
+	  debug_printf ("exception: '%s'\n", exception_msg.c_str ());
+	}
+
+      /* Report the failure.  */
+      SELF_CHECK (false);
+    }
+}
+
+/* Test the create_breakpoint_parse_arg_string function.  Just wraps
+   multiple calls to the test function above.  */
+
+static void
+create_breakpoint_parse_arg_string_tests (void)
+{
+  gdbarch *arch = current_inferior ()->arch ();
+  scoped_restore_current_pspace_and_thread restore;
+  scoped_mock_context<test_target_ops> mock_target (arch);
+
+  int global_thread_num = mock_target.mock_thread.global_num;
+
+  test ("  if blah  ", "blah");
+  test (" if blah thread 1", "blah", global_thread_num);
+  test (" if blah inferior 1", "blah", -1, 1);
+  test (" if blah thread 1  ", "blah", global_thread_num);
+  test ("thread 1 woof", nullptr, global_thread_num, -1, -1, false, "woof");
+  test ("thread 1 X", nullptr, global_thread_num, -1, -1, false, "X");
+  test (" if blah thread 1 -force-condition", "blah", global_thread_num,
+	-1, -1, true);
+  test (" -force-condition if blah thread 1", "blah", global_thread_num,
+	-1, -1, true);
+  test (" -force-condition if blah thread 1  ", "blah", global_thread_num,
+	-1, -1, true);
+  test ("thread 1 -force-condition if blah", "blah", global_thread_num,
+	-1, -1, true);
+  test ("if (A::outer::func ())", "(A::outer::func ())");
+}
+
+} // namespace selftests
+#endif /* GDB_SELF_TEST */
+
+void _initialize_break_cond_parse ();
+void
+_initialize_break_cond_parse ()
+{
+#if GDB_SELF_TEST
+  selftests::register_test
+    ("create_breakpoint_parse_arg_string",
+     selftests::create_breakpoint_parse_arg_string_tests);
+#endif
+}
diff --git a/gdb/break-cond-parse.h b/gdb/break-cond-parse.h
new file mode 100644
index 00000000000..b08b6fc6cc9
--- /dev/null
+++ b/gdb/break-cond-parse.h
@@ -0,0 +1,49 @@ 
+/* Copyright (C) 2023 Free Software Foundation, Inc.
+
+   This file is part of GDB.
+
+   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/>.  */
+
+#if !defined (BREAK_COND_PARSE_H)
+#define BREAK_COND_PARSE_H 1
+
+/* Given TOK, a string possibly containing a condition, thread, inferior,
+   task and force-condition flag, as accepted by the 'break' command,
+   extract the condition string, thread, inferior, task number, and the
+   force_condition flag, then set *COND_STRING, *THREAD, *INFERIOR, *TASK,
+   and *FORCE.
+
+   As TOK is parsed, if an unknown keyword is encountered before the 'if'
+   keyword then everything starting from the unknown keyword is placed into
+   *REST.
+
+   Both *COND and *REST are initialized to nullptr.  If no 'if' keyword is
+   found then *COND will be returned as nullptr.  If no unknown content is
+   found then *REST is returned as nullptr.
+
+   If no thread is found, *THREAD is set to -1.  If no inferior is found,
+   *INFERIOR is set to -1.  If no task is found, *TASK is set to -1.  If
+   the -force-condition flag is not found then *FORCE is set to false.
+
+   Due to the free-form nature that the string TOK might take (a 'thread'
+   keyword can appear before or after an 'if' condition) then we end up
+   having to check for keywords from both the start of TOK and the end of
+   TOK.  */
+
+extern void create_breakpoint_parse_arg_string
+  (const char *tok, gdb::unique_xmalloc_ptr<char> *cond_string,
+   int *thread, int *inferior, int *task,
+   gdb::unique_xmalloc_ptr<char> *rest, bool *force);
+
+#endif
diff --git a/gdb/breakpoint.c b/gdb/breakpoint.c
index 3ee23af83d6..e2cd4d360f2 100644
--- a/gdb/breakpoint.c
+++ b/gdb/breakpoint.c
@@ -70,6 +70,7 @@ 
 #include "cli/cli-style.h"
 #include "cli/cli-decode.h"
 #include <unordered_set>
+#include "break-cond-parse.h"
 
 /* readline include files */
 #include "readline/tilde.h"
@@ -6335,20 +6336,7 @@  print_breakpoint_location (const breakpoint *b, const bp_location *loc)
       uiout->field_stream ("at", stb);
     }
   else
-    {
-      uiout->field_string ("pending", b->locspec->to_string ());
-      /* If extra_string is available, it could be holding a condition
-	 or dprintf arguments.  In either case, make sure it is printed,
-	 too, but only for non-MI streams.  */
-      if (!uiout->is_mi_like_p () && b->extra_string != NULL)
-	{
-	  if (b->type == bp_dprintf)
-	    uiout->text (",");
-	  else
-	    uiout->text (" ");
-	  uiout->text (b->extra_string.get ());
-	}
-    }
+    uiout->field_string ("pending", b->locspec->to_string ());
 
   if (loc && is_breakpoint (b)
       && breakpoint_condition_evaluation_mode () == condition_evaluation_target
@@ -8713,8 +8701,8 @@  code_breakpoint::code_breakpoint (struct gdbarch *gdbarch_,
       gdb_assert (extra_string != nullptr);
       update_dprintf_command_list (this);
     }
-  else if (extra_string != nullptr)
-    error (_("Garbage '%s' at end of command"), extra_string.get ());
+  else
+    gdb_assert (extra_string == nullptr);
 
   /* The order of the locations is now stable.  Set the location
      condition using the location's number.  */
@@ -8942,197 +8930,6 @@  check_fast_tracepoint_sals (struct gdbarch *gdbarch,
     }
 }
 
-/* Given TOK, a string specification of condition and thread, as accepted
-   by the 'break' command, extract the condition string into *COND_STRING.
-   If no condition string is found then *COND_STRING is set to nullptr.
-
-   If the breakpoint specification has an associated thread, task, or
-   inferior, these are extracted into *THREAD, *TASK, and *INFERIOR
-   respectively, otherwise these arguments are set to -1 (for THREAD and
-   INFERIOR) or 0 (for TASK).
-
-   PC identifies the context at which the condition should be parsed.  */
-
-static void
-find_condition_and_thread (const char *tok, CORE_ADDR pc,
-			   gdb::unique_xmalloc_ptr<char> *cond_string,
-			   int *thread, int *inferior, int *task,
-			   gdb::unique_xmalloc_ptr<char> *rest)
-{
-  cond_string->reset ();
-  *thread = -1;
-  *inferior = -1;
-  *task = -1;
-  rest->reset ();
-  bool force = false;
-
-  while (tok && *tok)
-    {
-      const char *end_tok;
-      int toklen;
-      const char *cond_start = NULL;
-      const char *cond_end = NULL;
-
-      tok = skip_spaces (tok);
-
-      if ((*tok == '"' || *tok == ',') && rest)
-	{
-	  rest->reset (savestring (tok, strlen (tok)));
-	  break;
-	}
-
-      end_tok = skip_to_space (tok);
-
-      toklen = end_tok - tok;
-
-      if (toklen >= 1 && strncmp (tok, "if", toklen) == 0)
-	{
-	  tok = cond_start = end_tok + 1;
-	  try
-	    {
-	      parse_exp_1 (&tok, pc, block_for_pc (pc), 0);
-	    }
-	  catch (const gdb_exception_error &)
-	    {
-	      if (!force)
-		throw;
-	      else
-		tok = tok + strlen (tok);
-	    }
-	  cond_end = tok;
-	  cond_string->reset (savestring (cond_start, cond_end - cond_start));
-	}
-      else if (toklen >= 1 && strncmp (tok, "-force-condition", toklen) == 0)
-	{
-	  tok = tok + toklen;
-	  force = true;
-	}
-      else if (toklen >= 1 && strncmp (tok, "thread", toklen) == 0)
-	{
-	  const char *tmptok;
-	  struct thread_info *thr;
-
-	  if (*thread != -1)
-	    error(_("You can specify only one thread."));
-
-	  if (*task != -1)
-	    error (_("You can specify only one of thread or task."));
-
-	  if (*inferior != -1)
-	    error (_("You can specify only one of inferior or thread."));
-
-	  tok = end_tok + 1;
-	  thr = parse_thread_id (tok, &tmptok);
-	  if (tok == tmptok)
-	    error (_("Junk after thread keyword."));
-	  *thread = thr->global_num;
-	  tok = tmptok;
-	}
-      else if (toklen >= 1 && strncmp (tok, "inferior", toklen) == 0)
-	{
-	  if (*inferior != -1)
-	    error(_("You can specify only one inferior."));
-
-	  if (*task != -1)
-	    error (_("You can specify only one of inferior or task."));
-
-	  if (*thread != -1)
-	    error (_("You can specify only one of inferior or thread."));
-
-	  char *tmptok;
-	  tok = end_tok + 1;
-	  *inferior = strtol (tok, &tmptok, 0);
-	  if (tok == tmptok)
-	    error (_("Junk after inferior keyword."));
-	  if (!valid_global_inferior_id (*inferior))
-	    error (_("Unknown inferior number %d."), *inferior);
-	  tok = tmptok;
-	}
-      else if (toklen >= 1 && strncmp (tok, "task", toklen) == 0)
-	{
-	  char *tmptok;
-
-	  if (*task != -1)
-	    error(_("You can specify only one task."));
-
-	  if (*thread != -1)
-	    error (_("You can specify only one of thread or task."));
-
-	  if (*inferior != -1)
-	    error (_("You can specify only one of inferior or task."));
-
-	  tok = end_tok + 1;
-	  *task = strtol (tok, &tmptok, 0);
-	  if (tok == tmptok)
-	    error (_("Junk after task keyword."));
-	  if (!valid_task_id (*task))
-	    error (_("Unknown task %d."), *task);
-	  tok = tmptok;
-	}
-      else if (rest)
-	{
-	  rest->reset (savestring (tok, strlen (tok)));
-	  break;
-	}
-      else
-	error (_("Junk at end of arguments."));
-    }
-}
-
-/* Call 'find_condition_and_thread' for each sal in SALS until a parse
-   succeeds.  The parsed values are written to COND_STRING, THREAD,
-   TASK, and REST.  See the comment of 'find_condition_and_thread'
-   for the description of these parameters and INPUT.  */
-
-static void
-find_condition_and_thread_for_sals (const std::vector<symtab_and_line> &sals,
-				    const char *input,
-				    gdb::unique_xmalloc_ptr<char> *cond_string,
-				    int *thread, int *inferior, int *task,
-				    gdb::unique_xmalloc_ptr<char> *rest)
-{
-  int num_failures = 0;
-  for (auto &sal : sals)
-    {
-      gdb::unique_xmalloc_ptr<char> cond;
-      int thread_id = -1;
-      int inferior_id = -1;
-      int task_id = -1;
-      gdb::unique_xmalloc_ptr<char> remaining;
-
-      /* Here we want to parse 'arg' to separate condition from thread
-	 number.  But because parsing happens in a context and the
-	 contexts of sals might be different, try each until there is
-	 success.  Finding one successful parse is sufficient for our
-	 goal.  When setting the breakpoint we'll re-parse the
-	 condition in the context of each sal.  */
-      try
-	{
-	  find_condition_and_thread (input, sal.pc, &cond, &thread_id,
-				     &inferior_id, &task_id, &remaining);
-	  *cond_string = std::move (cond);
-	  /* A value of -1 indicates that these fields are unset.  At most
-	     one of these fields should be set (to a value other than -1)
-	     at this point.  */
-	  gdb_assert (((thread_id == -1 ? 1 : 0)
-		       + (task_id == -1 ? 1 : 0)
-		       + (inferior_id == -1 ? 1 : 0)) >= 2);
-	  *thread = thread_id;
-	  *inferior = inferior_id;
-	  *task = task_id;
-	  *rest = std::move (remaining);
-	  break;
-	}
-      catch (const gdb_exception_error &e)
-	{
-	  num_failures++;
-	  /* If no sal remains, do not continue.  */
-	  if (num_failures == sals.size ())
-	    throw;
-	}
-    }
-}
-
 /* Decode a static tracepoint marker spec.  */
 
 static std::vector<symtab_and_line>
@@ -9250,6 +9047,46 @@  create_breakpoint (struct gdbarch *gdbarch,
 	      || (type_wanted != bp_dprintf
 		  && (extra_string == nullptr || parse_extra)));
 
+  /* Will hold either copies of the similarly named function argument, or
+     will hold a modified version of the function argument, depending on
+     the value of PARSE_EXTRA.  */
+  gdb::unique_xmalloc_ptr<char> cond_string_copy;
+  gdb::unique_xmalloc_ptr<char> extra_string_copy;
+
+  if (parse_extra)
+    {
+      /* Parse EXTRA_STRING splitting the parts out.  */
+      create_breakpoint_parse_arg_string (extra_string, &cond_string_copy,
+					  &thread, &inferior, &task,
+					  &extra_string_copy,
+					  &force_condition);
+
+      /* We could check that EXTRA_STRING_COPY is empty at this point -- it
+	 should be, as we only get here for things that are not bp_dprintf,
+	 however, we prefer to give the location spec parser a chance to
+	 run first, this means the user will get errors about invalid
+	 location spec instead of an error about garbage at the end of the
+	 command line.
+
+	 We still do the EXTRA_STRING_COPY is empty check, just later in
+	 this function.  */
+
+      gdb_assert (thread == -1 || thread > 0);
+      gdb_assert (task == -1 || task > 0);
+      gdb_assert (inferior == -1 || inferior > 0);
+    }
+  else
+    {
+      if (cond_string != nullptr)
+	cond_string_copy.reset (xstrdup (cond_string));
+      if (extra_string != nullptr)
+	extra_string_copy.reset (xstrdup (extra_string));
+    }
+
+  /* Clear these.  Updated values are now held in the *_copy locals.  */
+  cond_string = nullptr;
+  extra_string = nullptr;
+
   try
     {
       ops->create_sals_from_location_spec (locspec, &canonical);
@@ -9285,6 +9122,13 @@  create_breakpoint (struct gdbarch *gdbarch,
 	throw;
     }
 
+  /* The only bp_dprintf should have anything in EXTRA_STRING_COPY by this
+     point.  For all other breakpoints this indicates an error.  We could
+     place this check earlier in the function, but we prefer to see errors
+     from the location spec parser before we see this error message.  */
+  if (type_wanted != bp_dprintf && extra_string_copy.get () != nullptr)
+    error (_("Garbage '%s' at end of command"), extra_string_copy.get ());
+
   if (!pending && canonical.lsals.empty ())
     return 0;
 
@@ -9308,63 +9152,31 @@  create_breakpoint (struct gdbarch *gdbarch,
      breakpoint.  */
   if (!pending)
     {
-      gdb::unique_xmalloc_ptr<char> cond_string_copy;
-      gdb::unique_xmalloc_ptr<char> extra_string_copy;
-
-      if (parse_extra)
+      /* Check the validity of the condition.  We should error out if the
+	 condition is invalid at all of the locations and if it is not
+	 forced.  In the PARSE_EXTRA case above, this check is done when
+	 parsing the EXTRA_STRING.  */
+      if (cond_string_copy.get () != nullptr && !force_condition)
 	{
-	  gdb_assert (type_wanted != bp_dprintf);
-
-	  gdb::unique_xmalloc_ptr<char> rest;
-	  gdb::unique_xmalloc_ptr<char> cond;
-
+	  int num_failures = 0;
 	  const linespec_sals &lsal = canonical.lsals[0];
-
-	  find_condition_and_thread_for_sals (lsal.sals, extra_string,
-					      &cond, &thread, &inferior,
-					      &task, &rest);
-
-	  if (rest.get () != nullptr && *(rest.get ()) != '\0')
-	    error (_("Garbage '%s' at end of command"), rest.get ());
-
-	  cond_string_copy = std::move (cond);
-	  extra_string_copy = std::move (rest);
-	}
-      else
-	{
-	  /* Check the validity of the condition.  We should error out
-	     if the condition is invalid at all of the locations and
-	     if it is not forced.  In the PARSE_EXTRA case above, this
-	     check is done when parsing the EXTRA_STRING.  */
-	  if (cond_string != nullptr && !force_condition)
+	  for (const auto &sal : lsal.sals)
 	    {
-	      int num_failures = 0;
-	      const linespec_sals &lsal = canonical.lsals[0];
-	      for (const auto &sal : lsal.sals)
+	      const char *cond = cond_string_copy.get ();
+	      try
 		{
-		  const char *cond = cond_string;
-		  try
-		    {
-		      parse_exp_1 (&cond, sal.pc, block_for_pc (sal.pc), 0);
-		      /* One success is sufficient to keep going.  */
-		      break;
-		    }
-		  catch (const gdb_exception_error &)
-		    {
-		      num_failures++;
-		      /* If this is the last sal, error out.  */
-		      if (num_failures == lsal.sals.size ())
-			throw;
-		    }
+		  parse_exp_1 (&cond, sal.pc, block_for_pc (sal.pc), 0);
+		  /* One success is sufficient to keep going.  */
+		  break;
+		}
+	      catch (const gdb_exception_error &)
+		{
+		  num_failures++;
+		  /* If this is the last sal, error out.  */
+		  if (num_failures == lsal.sals.size ())
+		    throw;
 		}
 	    }
-
-	  /* Create a private copy of condition string.  */
-	  if (cond_string)
-	    cond_string_copy.reset (xstrdup (cond_string));
-	  /* Create a private copy of any extra string.  */
-	  if (extra_string)
-	    extra_string_copy.reset (xstrdup (extra_string));
 	}
 
       ops->create_breakpoints_sal (gdbarch, &canonical,
@@ -9381,21 +9193,16 @@  create_breakpoint (struct gdbarch *gdbarch,
 								 type_wanted);
       b->locspec = locspec->clone ();
 
-      if (parse_extra)
-	b->cond_string = NULL;
-      else
-	{
-	  /* Create a private copy of condition string.  */
-	  b->cond_string.reset (cond_string != NULL
-				? xstrdup (cond_string)
-				: NULL);
-	  b->thread = thread;
-	}
+      /* Create a private copy of the condition string.  */
+      b->cond_string = std::move (cond_string_copy);
+
+      b->thread = thread;
+      b->task = task;
+      b->inferior = inferior;
 
       /* Create a private copy of any extra string.  */
-      b->extra_string.reset (extra_string != NULL
-			     ? xstrdup (extra_string)
-			     : NULL);
+      b->extra_string = std::move (extra_string_copy);
+
       b->ignore_count = ignore_count;
       b->disposition = tempflag ? disp_del : disp_donttouch;
       b->condition_not_parsed = 1;
@@ -9404,9 +9211,12 @@  create_breakpoint (struct gdbarch *gdbarch,
 	   && type_wanted != bp_hardware_breakpoint) || thread != -1)
 	b->pspace = current_program_space;
 
+      if (b->type == bp_dprintf)
+	update_dprintf_command_list (b.get ());
+
       install_breakpoint (internal, std::move (b), 0);
     }
-  
+
   if (canonical.lsals.size () > 1)
     {
       warning (_("Multiple breakpoints were set.\nUse the "
@@ -13151,24 +12961,6 @@  code_breakpoint::location_spec_to_sals (location_spec *locspec,
     {
       for (auto &sal : sals)
 	resolve_sal_pc (&sal);
-      if (condition_not_parsed && extra_string != NULL)
-	{
-	  gdb::unique_xmalloc_ptr<char> local_cond, local_extra;
-	  int local_thread, local_task, local_inferior;
-
-	  find_condition_and_thread_for_sals (sals, extra_string.get (),
-					      &local_cond, &local_thread,
-					      &local_inferior,
-					      &local_task, &local_extra);
-	  gdb_assert (cond_string == nullptr);
-	  if (local_cond != nullptr)
-	    cond_string = std::move (local_cond);
-	  thread = local_thread;
-	  task = local_task;
-	  if (local_extra != nullptr)
-	    extra_string = std::move (local_extra);
-	  condition_not_parsed = 0;
-	}
 
       if (type == bp_static_tracepoint)
 	{
diff --git a/gdb/testsuite/gdb.ada/tasks.exp b/gdb/testsuite/gdb.ada/tasks.exp
index 603d43f7c36..337ee1b6f07 100644
--- a/gdb/testsuite/gdb.ada/tasks.exp
+++ b/gdb/testsuite/gdb.ada/tasks.exp
@@ -55,11 +55,11 @@  gdb_test "watch j task 1 task 3" "You can specify only one task\\."
 
 # Check that attempting to combine 'task' and 'thread' gives an error.
 gdb_test "break break_me task 1 thread 1" \
-    "You can specify only one of thread or task\\."
+    "You can specify only one of thread, inferior, or task\\."
 gdb_test "break break_me thread 1 task 1" \
-    "You can specify only one of thread or task\\."
+    "You can specify only one of thread, inferior, or task\\."
 gdb_test "break break_me inferior 1 task 1" \
-    "You can specify only one of inferior or task\\."
+    "You can specify only one of thread, inferior, or task\\."
 gdb_test "watch j task 1 thread 1" \
     "You can specify only one of thread or task\\."
 gdb_test "watch j thread 1 task 1" \
diff --git a/gdb/testsuite/gdb.base/condbreak.exp b/gdb/testsuite/gdb.base/condbreak.exp
index a5b2a28701b..5a6a42e073b 100644
--- a/gdb/testsuite/gdb.base/condbreak.exp
+++ b/gdb/testsuite/gdb.base/condbreak.exp
@@ -179,6 +179,10 @@  gdb_test "break -q main if (1==1) thread 999" \
     "Unknown thread 999\\."
 gdb_test "break -q main thread 999 if (1==1)" \
     "Unknown thread 999\\."
+gdb_test "break -q main if (1==1) thread 999 -force-condition" \
+    "Unknown thread 999\\."
+gdb_test "break -q main thread 999 if (1==1) -force-condition" \
+    "Unknown thread 999\\."
 
 # Verify that both if and thread can be distinguished from a breakpoint
 # address expression.
@@ -186,20 +190,71 @@  gdb_test "break *main if (1==1) thread 999" \
     "Unknown thread 999\\."
 gdb_test "break *main thread 999 if (1==1)" \
     "Unknown thread 999\\."
+gdb_test "break *main if (1==1) thread 999 -force-condition" \
+    "Unknown thread 999\\."
+gdb_test "break *main thread 999 if (1==1) -force-condition" \
+    "Unknown thread 999\\."
 
 # Similarly for task.
 gdb_test "break *main if (1==1) task 999" \
     "Unknown task 999\\."
 gdb_test "break *main task 999 if (1==1)" \
     "Unknown task 999\\."
+gdb_test "break *main if (1==1) task 999 -force-condition" \
+    "Unknown task 999\\."
+gdb_test "break *main task 999 if (1==1) -force-condition" \
+    "Unknown task 999\\."
 
-# GDB accepts abbreviations for "thread" and "task".
+# GDB accepts abbreviations for "thread", "task" and
+# "-force-condition", when these keywords appear after
+# the breakpoint condition.
 gdb_test "break *main if (1==1) t 999" \
     "Unknown thread 999\\."
 gdb_test "break *main if (1==1) th 999" \
     "Unknown thread 999\\."
 gdb_test "break *main if (1==1) ta 999" \
     "Unknown task 999\\."
+gdb_test "break *main if (1==1) t 999 -force" \
+    "Unknown thread 999\\."
+gdb_test "break *main if (1==1) th 999 -force" \
+    "Unknown thread 999\\."
+gdb_test "break *main if (1==1) ta 999 -force" \
+    "Unknown task 999\\."
+
+# Check the use of abbreviations before the condition.  This works
+# because, when the location spec starts with '*' GDB is able to
+# figure out that the complete location is '*main'.
+gdb_test "break *main t 999 if (1==1)" \
+    "Unknown thread 999\\."
+gdb_test "break *main th 999 if (1==1)" \
+    "Unknown thread 999\\."
+gdb_test "break *main ta 999 if (1==1)" \
+    "Unknown task 999\\."
+gdb_test "break *main t 999 -force if (1==1)" \
+    "Unknown thread 999\\."
+gdb_test "break *main th 999 -force if (1==1)" \
+    "Unknown thread 999\\."
+gdb_test "break *main ta 999 -force if (1==1)" \
+    "Unknown task 999\\."
+
+# However, when the location spec doesn't start with '*' GDB relies on
+# the linespec parser to spot the keyword which marks the end of the
+# linespec, and this parser doesn't check for abbreviations.
+gdb_test "with breakpoint pending off -- break main t 999 if (1==1)" \
+    "Function \"main t 999\" not defined\\."
+gdb_test "with breakpoint pending off -- break main th 999 if (1==1)" \
+    "Function \"main th 999\" not defined\\."
+gdb_test "with breakpoint pending off -- break main ta 999 if (1==1)" \
+    "Function \"main ta 999\" not defined\\."
+
+# GDB does not treat a "-force-condition" flag that appears
+# immediately after the condition as the flag, but instead treats it
+# as " - force - condition", that is, subtraction of the symbol
+# "force" followed by subtraction of symbol "context".  This is really
+# just a quirk of how this used to be implemented, and should maybe be
+# changed in the future.  However, for now GDB retains this behaviour.
+gdb_test "break *main if (1==1) -force-condition" \
+    "No symbol \"force\" in current context\\."
 
 set test "run until breakpoint at marker3"
 gdb_test_multiple "continue" $test {
diff --git a/gdb/testsuite/gdb.base/pending.exp b/gdb/testsuite/gdb.base/pending.exp
index d7d3735000a..645f63bdb16 100644
--- a/gdb/testsuite/gdb.base/pending.exp
+++ b/gdb/testsuite/gdb.base/pending.exp
@@ -170,7 +170,8 @@  gdb_test "info break" \
 \[\t \]+stop only if k == 1.*
 \[\t \]+print k.*
 \[0-9\]+\[\t \]+breakpoint     keep y.* in main at .*$srcfile:$mainline.*
-\[0-9\]+\[\t \]+breakpoint     keep y.*PENDING.*pendshr.c:$bp2_loc if x > 3.*" \
+\[0-9\]+\[\t \]+breakpoint     keep y.*PENDING.*pendshr.c:$bp2_loc.*
+\\s+stop only if x > 3.*" \
 "multiple pending breakpoints"
 
 
@@ -195,8 +196,10 @@  gdb_test "info break" \
 \[\t \]+stop only if k == 1.*
 \[\t \]+print k.*
 \[0-9\]+\[\t \]+breakpoint     keep y.* in main at .*$srcfile:$mainline.*
-\[0-9\]+\[\t \]+breakpoint     keep y.*PENDING.*pendshr.c:$bp2_loc if x > 3.*
-\[0-9\]+\[\t \]+breakpoint     keep y.*PENDING.*pendshr.c:$bp3_loc.*ignore next 2 hits.*" \
+\[0-9\]+\[\t \]+breakpoint     keep y.*PENDING.*pendshr.c:$bp2_loc.*
+\\s+stop only if x > 3.*
+\[0-9\]+\[\t \]+breakpoint     keep y.*PENDING.*pendshr.c:$bp3_loc.*
+\\s+ignore next 2 hits.*" \
 "multiple pending breakpoints 2"
 
 #
@@ -267,3 +270,17 @@  gdb_test "info break" \
 \[0-9\]+\[\t \]+breakpoint     keep y.* in main at .*$srcfile:$mainline.*
 \[0-9\]+\[\t \]+breakpoint     keep y.*PENDING.*imaginary.*" \
 "verify pending breakpoint after restart"
+
+# Test GDB's parsing of pending breakpoint thread and condition.
+
+gdb_test_no_output "set breakpoint pending on"
+gdb_test "break foo if (unknown_var && another_unknown_var) thread 1" \
+    "Breakpoint $decimal \\(foo\\) pending\\."
+set bpnum [get_integer_valueof "\$bpnum" "*INVALID" \
+	       "get number for foo breakpoint"]
+gdb_test "info breakpoints $bpnum" \
+    [multi_line \
+	 "$bpnum\\s+breakpoint\\s+keep\\s+y\\s+<PENDING>\\s+foo" \
+	 "\\s+stop only if \\(unknown_var && another_unknown_var\\)" \
+	 "\\s+stop only in thread 1"] \
+    "check pending breakpoint on foo"
diff --git a/gdb/testsuite/gdb.linespec/explicit.exp b/gdb/testsuite/gdb.linespec/explicit.exp
index 668002d9038..8cbefe204ec 100644
--- a/gdb/testsuite/gdb.linespec/explicit.exp
+++ b/gdb/testsuite/gdb.linespec/explicit.exp
@@ -575,7 +575,7 @@  namespace eval $testfile {
 	      allow-pending]} {
 	fail "set $tst"
     } else {
-	gdb_test "info break" ".*PENDING.*myfunction if foofoofoo == 1.*" $tst
+	gdb_test "info break" ".*PENDING.*myfunction\r\n\\s+stop only if foofoofoo == 1.*" $tst
     }
 
     gdb_exit
@@ -586,7 +586,7 @@  namespace eval $testfile {
 	      allow-pending]} {
 	fail "set $tst"
     } else {
-	gdb_test "info break" ".*PENDING.*myfunction if arg == 0" $tst
+	gdb_test "info break" ".*PENDING.*myfunction\r\n\\s+stop only if arg == 0" $tst
 
 	gdb_load [standard_output_file $exefile]
 	gdb_test "info break" \
diff --git a/gdb/testsuite/gdb.mi/mi-dprintf-pending.exp b/gdb/testsuite/gdb.mi/mi-dprintf-pending.exp
index 28f52938aeb..bb9987a5572 100644
--- a/gdb/testsuite/gdb.mi/mi-dprintf-pending.exp
+++ b/gdb/testsuite/gdb.mi/mi-dprintf-pending.exp
@@ -50,7 +50,8 @@  set bp_location1 [gdb_get_line_number "set breakpoint 1 here"]
 # Set pending dprintf via MI.
 set bp [mi_make_breakpoint_pending -number "1" -type "dprintf" \
 	    -disp "keep" -enabled "y" -pending "pendfunc1" \
-	    -original-location "pendfunc1"]
+	    -original-location "pendfunc1" \
+	    -script {\["printf \\\"hello\\\""\]}]
 mi_gdb_test "-dprintf-insert -f pendfunc1 \"hello\"" \
     ".*\\^done,$bp" "mi set dprintf"
 
diff --git a/gdb/testsuite/gdb.multi/inferior-specific-bp.exp b/gdb/testsuite/gdb.multi/inferior-specific-bp.exp
index 1f6573268da..b8aceabcad6 100644
--- a/gdb/testsuite/gdb.multi/inferior-specific-bp.exp
+++ b/gdb/testsuite/gdb.multi/inferior-specific-bp.exp
@@ -51,9 +51,9 @@  if {![runto_main]} {
 # this should fail.  Try with the keywords in both orders just in case the
 # parser has a bug.
 gdb_test "break foo thread 1.1 inferior 1" \
-    "You can specify only one of inferior or thread\\."
+    "You can specify only one of thread, inferior, or task\\."
 gdb_test "break foo inferior 1 thread 1.1" \
-    "You can specify only one of inferior or thread\\."
+    "You can specify only one of thread, inferior, or task\\."
 
 # Try to create a breakpoint using the 'inferior' keyword multiple times.
 gdb_test "break foo inferior 1 inferior 2" \
diff --git a/gdb/testsuite/gdb.threads/del-pending-thread-bp-lib.c b/gdb/testsuite/gdb.threads/del-pending-thread-bp-lib.c
new file mode 100644
index 00000000000..15d1b9833dd
--- /dev/null
+++ b/gdb/testsuite/gdb.threads/del-pending-thread-bp-lib.c
@@ -0,0 +1,22 @@ 
+/* Copyright 2023 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/>.  */
+
+int global_var = 0;
+
+void
+foo (int arg)
+{
+  global_var = arg;
+}
diff --git a/gdb/testsuite/gdb.threads/del-pending-thread-bp.c b/gdb/testsuite/gdb.threads/del-pending-thread-bp.c
new file mode 100644
index 00000000000..938e05bc4d7
--- /dev/null
+++ b/gdb/testsuite/gdb.threads/del-pending-thread-bp.c
@@ -0,0 +1,85 @@ 
+/* Copyright 2023 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/>.  */
+
+#include <dlfcn.h>
+#include <pthread.h>
+#include <stdlib.h>
+
+pthread_barrier_t barrier;
+
+static void
+barrier_wait (pthread_barrier_t *b)
+{
+  int res = pthread_barrier_wait (b);
+  if (res != 0 && res != PTHREAD_BARRIER_SERIAL_THREAD)
+    abort ();
+}
+
+static void *
+thread_worker (void *arg)
+{
+  barrier_wait (&barrier);
+  return NULL;
+}
+
+void
+breakpt ()
+{
+  /* Nothing.  */
+}
+
+int
+main (void)
+{
+  void *handle;
+  void (*func)(int);
+  pthread_t thread;
+
+  if (pthread_barrier_init (&barrier, NULL, 2) != 0)
+    abort ();
+
+  if (pthread_create (&thread, NULL, thread_worker, NULL) != 0)
+    abort ();
+
+  breakpt ();
+
+  /* Allow the worker thread to complete.  */
+  barrier_wait (&barrier);
+
+  if (pthread_join (thread, NULL) != 0)
+    abort ();
+
+  breakpt ();
+
+  /* Now load the shared library.  */
+  handle = dlopen (SHLIB_NAME, RTLD_LAZY);
+  if (handle == NULL)
+    abort ();
+
+  /* Find the function symbol.  */
+  func = (void (*)(int)) dlsym (handle, "foo");
+
+  /* Call the library function.  */
+  func (1);
+
+  /* Unload the shared library.  */
+  if (dlclose (handle) != 0)
+    abort ();
+
+  breakpt ();
+
+  return 0;
+}
+
diff --git a/gdb/testsuite/gdb.threads/del-pending-thread-bp.exp b/gdb/testsuite/gdb.threads/del-pending-thread-bp.exp
new file mode 100644
index 00000000000..4ed05eaa42d
--- /dev/null
+++ b/gdb/testsuite/gdb.threads/del-pending-thread-bp.exp
@@ -0,0 +1,108 @@ 
+# Copyright 2023 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/>.
+
+# This test checks that pending thread-specific breakpoints are
+# correctly deleted when the thread the breakpoint is for goes out of
+# scope.
+#
+# We also check that we can't create a pending thread-specific
+# breakpoint for a non-existent thread.
+
+require allow_shlib_tests
+
+standard_testfile
+
+set libname $testfile-lib
+set srcfile_lib $srcdir/$subdir/$libname.c
+set binfile_lib [standard_output_file $libname.so]
+
+if { [gdb_compile_shlib $srcfile_lib $binfile_lib {}] != "" } {
+    untested "failed to compile shared library 1"
+    return -1
+}
+
+set binfile_lib_target [gdb_download_shlib $binfile_lib]
+
+if { [prepare_for_testing "failed to prepare" $testfile $srcfile \
+	  [list debug \
+	       additional_flags=-DSHLIB_NAME=\"$binfile_lib_target\" \
+	       shlib_load pthreads]] } {
+    return -1
+}
+
+gdb_locate_shlib $binfile_lib
+
+if ![runto_main] {
+    return 0
+}
+
+# Run until we have two threads.
+gdb_breakpoint "breakpt"
+gdb_continue_to_breakpoint "first breakpt call"
+
+# Confirm that we have a thread '2'.
+gdb_test "info threads" "\r\n\\s+2\\s+\[^\r\n\]+"
+
+# Create a pending, thread-specific, breakpoint on 'foo'.
+gdb_breakpoint "foo thread 2" allow-pending
+set bpnum [get_integer_valueof "\$bpnum" "*INVALID*" \
+	       "get breakpoint number"]
+
+# Check we can't create a pending thread-specific breakpoint for a
+# non-existent thread.
+gdb_test "with breakpoint pending on -- break foo thread 99" \
+    "Unknown thread 99\\."
+
+# Continue to 'breakpt' again.  Don't use gdb_continue_to_breakpoint
+# as we are looking for the thread exited and breakpoint deleted
+# messages.
+set output [list "Continuing\\."]
+
+if {!([target_info exists gdb_protocol]
+      && ([target_info gdb_protocol] == "remote"
+	  || [target_info gdb_protocol] == "extended-remote"))} {
+    # Due to bug PR gdb/30129 we don't see the thread exited messages
+    # for remote targets.  When this bit of this test can be cleaned
+    # up.
+    lappend output "\\\[Thread \[^\r\n\]+ exited\\\]"
+}
+
+lappend output "Thread-specific breakpoint $bpnum deleted - thread 2 no longer in the thread list\\." \
+    "" \
+    "Thread 1 \[^\r\n\]+, breakpt \\(\\) at \[^\r\n\]+" \
+    "$decimal\\s+\[^\r\n\]+"
+
+gdb_test "continue" \
+    [multi_line {*}$output] \
+    "second breakpt call"
+
+# Confirm breakpoint has been deleted.
+gdb_test "info breakpoints $bpnum" \
+    "No breakpoint or watchpoint matching '$bpnum'\\."
+
+# Continue again.  This will pass through 'foo'.  We should not stop
+# in 'foo', the breakpoint has been deleted.  We should next stop in
+# breakpt again.
+gdb_test "continue" \
+    [multi_line \
+	 "Continuing\\." \
+	 "" \
+	 "Thread 1 \[^\r\n\]+ hit Breakpoint $decimal, breakpt \\(\\) at \[^\r\n\]+" \
+	 "$decimal\\s+\[^\r\n\]+"] \
+    "third breakpt call"
+gdb_test "bt 1" \
+    [multi_line \
+	 "#0\\s+breakpt \\(\\) at \[^\r\n\]+" \
+	 "\\(More stack frames follow\\.\\.\\.\\)"]