[PATCH 0/2] Tab complete convenience variables

[PATCH 0/2] Tab complete convenience variables

[Antonio Rische]

Implement tab completion for convenience variables. This has been
something that annoyed me about gdb for years.

Antonio Rische (2):
  gdb: Do not create variables when parsing expressions
  gdb: Tab complete internalvars in expressions

 gdb/ax-gdb.c        | 25 +++++++++++++++++++++++
 gdb/cli/cli-utils.c |  2 +-
 gdb/completer.c     | 10 ++++++++++
 gdb/expop.h         | 48 ++++++++++++++++++++++++++++++++++++++++++++-
 gdb/parse.c         |  3 +--
 5 files changed, 84 insertions(+), 4 deletions(-)

-- 
2.46.0

[PATCH 1/2] gdb: Do not create variables when parsing expressions

[Antonio Rische]

This is not particularly useful on its own, but allows us to parse
expressions without having side-effects on the set of internalvars.

This will then allow us to tab-complete internalvars in expressions
without creating new internalvars, which would otherwise defeat the
purpose of tab completion (completing only variables that have been
set).
---
 gdb/ax-gdb.c        | 25 +++++++++++++++++++++++
 gdb/cli/cli-utils.c |  2 +-
 gdb/expop.h         | 48 ++++++++++++++++++++++++++++++++++++++++++++-
 gdb/parse.c         |  3 +--
 4 files changed, 74 insertions(+), 4 deletions(-)

diff --git a/gdb/ax-gdb.c b/gdb/ax-gdb.c
index 3d619012b..7853aa444 100644
--- a/gdb/ax-gdb.c
+++ b/gdb/ax-gdb.c
@@ -1668,6 +1668,31 @@ register_operation::do_generate_ax (struct expression *exp,
   value->type = register_type (ax->gdbarch, reg);
 }
 
+void
+uninit_internalvar_operation::do_generate_ax (struct expression *exp,
+				       struct agent_expr *ax,
+				       struct axs_value *value,
+				       struct type *cast_type)
+{
+  const char *name = get_name ();
+  struct internalvar *var = lookup_only_internalvar(name);
+  struct trace_state_variable *tsv;
+
+  tsv = find_trace_state_variable (name);
+  if (tsv)
+    {
+      ax_tsv (ax, aop_getv, tsv->number);
+      if (ax->tracing)
+	ax_tsv (ax, aop_tracev, tsv->number);
+      /* Trace state variables are always 64-bit integers.  */
+      value->kind = axs_rvalue;
+      value->type = builtin_type (ax->gdbarch)->builtin_long_long;
+    }
+  else if (! compile_internalvar_to_ax (var, ax, value))
+    error (_("$%s is not a trace state variable; GDB agent "
+	     "expressions cannot use convenience variables."), name);
+}
+
 void
 internalvar_operation::do_generate_ax (struct expression *exp,
 				       struct agent_expr *ax,
diff --git a/gdb/cli/cli-utils.c b/gdb/cli/cli-utils.c
index 45b30842e..7dedb3829 100644
--- a/gdb/cli/cli-utils.c
+++ b/gdb/cli/cli-utils.c
@@ -49,7 +49,7 @@ get_ulongest (const char **pp, int trailer)
 	  while (isalnum (*p) || *p == '_')
 	    p++;
 	  std::string varname (start, p - start);
-	  if (!get_internalvar_integer (lookup_internalvar (varname.c_str ()),
+	  if (!get_internalvar_integer (lookup_only_internalvar (varname.c_str ()),
 				       &retval))
 	    error (_("Convenience variable $%s does not have integer value."),
 		   varname.c_str ());
diff --git a/gdb/expop.h b/gdb/expop.h
index 2d46a9dad..f0cc8dbb1 100644
--- a/gdb/expop.h
+++ b/gdb/expop.h
@@ -878,6 +878,42 @@ class bool_operation
   { return true; }
 };
 
+/* Reference to a variable that had not been defined at parse time.  */
+class uninit_internalvar_operation
+  : public tuple_holding_operation<std::string>
+{
+public:
+
+  using tuple_holding_operation::tuple_holding_operation;
+
+  value *evaluate (struct type *expect_type,
+		   struct expression *exp,
+		   enum noside noside) override
+  {
+    internalvar* iv = lookup_only_internalvar(std::get<0> (m_storage).c_str ());
+    if (!iv) {
+      return value::allocate (builtin_type (exp->gdbarch)->builtin_void);
+    }
+    return value_of_internalvar (exp->gdbarch, iv);
+  }
+
+  const char *get_name () const
+  {
+    return std::get<0> (m_storage).c_str ();
+  }
+
+  enum exp_opcode opcode () const override
+  { return OP_INTERNALVAR; }
+
+protected:
+
+  void do_generate_ax (struct expression *exp,
+		       struct agent_expr *ax,
+		       struct axs_value *value,
+		       struct type *cast_type)
+    override;
+};
+
 class internalvar_operation
   : public tuple_holding_operation<internalvar *>
 {
@@ -1891,7 +1927,17 @@ class assign_operation
 		   struct expression *exp,
 		   enum noside noside) override
   {
-    value *lhs = std::get<0> (m_storage)->evaluate (nullptr, exp, noside);
+    operation *lhs_op = std::get<0> (m_storage).get();
+
+    /* If the operation is an internalvar that was not initialized at
+       parse time, ensure that it exists. */
+    auto *uninit_var_op = dynamic_cast<uninit_internalvar_operation *> (lhs_op);
+    if (uninit_var_op) {
+      lookup_internalvar (uninit_var_op->get_name());
+    }
+
+    value *lhs = lhs_op->evaluate (nullptr, exp, noside);
+
     /* Special-case assignments where the left-hand-side is a
        convenience variable -- in these, don't bother setting an
        expected type.  This avoids a weird case where re-assigning a
diff --git a/gdb/parse.c b/gdb/parse.c
index e0837de7b..19e2504f9 100644
--- a/gdb/parse.c
+++ b/gdb/parse.c
@@ -242,8 +242,7 @@ parser_state::push_dollar (struct stoken str)
 
   /* Any other names are assumed to be debugger internal variables.  */
 
-  push_new<expr::internalvar_operation>
-    (create_internalvar (copy.c_str () + 1));
+  push_new<expr::uninit_internalvar_operation> (copy.c_str () + 1);
 }
 
 /* See parser-defs.h.  */
-- 
2.46.0

[PATCH 2/2] gdb: Tab complete internalvars in expressions

[Antonio Rische]

For example, 'print $f<tab>' after running 'set $foo=0' now tab
completes.

'print $_siginf<tab>' also now tab completes.
---
 gdb/completer.c | 10 ++++++++++
 1 file changed, 10 insertions(+)

diff --git a/gdb/completer.c b/gdb/completer.c
index 1008ec23b..6476ddf79 100644
--- a/gdb/completer.c
+++ b/gdb/completer.c
@@ -1099,6 +1099,16 @@ complete_expression (completion_tracker &tracker,
       && expr_completer->complete (exp.get (), tracker))
     return;
 
+  if (text[0] == '$')
+	{
+	  tracker.advance_custom_word_point_by (1);
+	  /* We don't support completion of history indices.  */
+	  if (!isdigit (text[1]))
+	    complete_internalvar (tracker, &text[1]);
+	  tracker.advance_custom_word_point_by (-1);
+	  return;
+	}
+
   complete_files_symbols (tracker, text, word);
 }
 
-- 
2.46.0

Re: [PATCH 2/2] gdb: Tab complete internalvars in expressions

[nt8r]

It seems I misunderstood the semantics of the 'word' and 'text' parameters here. I'll correct in v2; as written this doesn't work for completions not at the start of the text to complete (e.g. 'print 4+$fo<tab>' or 'print $fo<tab>+4').

On Thursday, August 22nd, 2024 at 3:47 PM, Antonio Rische <nt8r@protonmail.com> wrote:

> For example, 'print $f<tab>' after running 'set $foo=0' now tab
> 
> completes.
> 
> 'print $_siginf<tab>' also now tab completes.
> 
> ---
> gdb/completer.c | 10 ++++++++++
> 1 file changed, 10 insertions(+)
> 
> diff --git a/gdb/completer.c b/gdb/completer.c
> index 1008ec23b..6476ddf79 100644
> --- a/gdb/completer.c
> +++ b/gdb/completer.c
> @@ -1099,6 +1099,16 @@ complete_expression (completion_tracker &tracker,
> && expr_completer->complete (exp.get (), tracker))
> 
> return;
> 
> + if (text[0] == '$')
> + {
> + tracker.advance_custom_word_point_by (1);
> + /* We don't support completion of history indices. */
> + if (!isdigit (text[1]))
> + complete_internalvar (tracker, &text[1]);
> + tracker.advance_custom_word_point_by (-1);
> + return;
> + }
> +
> complete_files_symbols (tracker, text, word);
> }
> 
> --
> 2.46.0

Re: [PATCH 2/2] gdb: Tab complete internalvars in expressions

[Keith Seitz]

Hi,

On 8/22/24 12:41 PM, nt8r@protonmail.com wrote:
> It seems I misunderstood the semantics of the 'word' and 'text' parameters here. I'll correct in v2; as written this doesn't work for completions not at the start of the text to complete (e.g. 'print 4+$fo<tab>' or 'print $fo<tab>+4').

Wow, thank you for taking a look at this! This is a most welcome
submission.

Is it possible to include any tests exercising this new feature? That
would make review much easier for us.

There are several examples in gdb/testsuite/ to follow. Grep
for "test_gdb_complete" or take a look at the support routines
in gdb/testsuite/lib/completion-support.exp. [Also see
https://sourceware.org/gdb/wiki/GDBTestcaseCookbook and
gdb/testsuite/README for add'l pointers/examples on writing
and running tests.]

Do you have an assignment on file? If not, in order for us to
accept your patch, please see/follow

https://sourceware.org/gdb/wiki/ContributionChecklist#FSF_copyright_Assignment

Thank you again for submitting a patch to fix this.

Keith

Re: [PATCH 2/2] gdb: Tab complete internalvars in expressions

[nt8r]

Thanks for the enthusiasm! I'm adding some tests in v2.

How do I go about copyright assignment? The documents I'm seeing seem to suggest getting in touch with a maintainer over e-mail.

Thanks,
Antonio

On Monday, August 26th, 2024 at 4:54 PM, Keith Seitz <keiths@redhat.com> wrote:

> Hi,
> 
> On 8/22/24 12:41 PM, nt8r@protonmail.com wrote:
> 
> > It seems I misunderstood the semantics of the 'word' and 'text' parameters here. I'll correct in v2; as written this doesn't work for completions not at the start of the text to complete (e.g. 'print 4+$fo<tab>' or 'print $fo<tab>+4').
> 
> 
> Wow, thank you for taking a look at this! This is a most welcome
> submission.
> 
> Is it possible to include any tests exercising this new feature? That
> would make review much easier for us.
> 
> There are several examples in gdb/testsuite/ to follow. Grep
> for "test_gdb_complete" or take a look at the support routines
> in gdb/testsuite/lib/completion-support.exp. [Also see
> https://sourceware.org/gdb/wiki/GDBTestcaseCookbook and
> gdb/testsuite/README for add'l pointers/examples on writing
> and running tests.]
> 
> Do you have an assignment on file? If not, in order for us to
> accept your patch, please see/follow
> 
> https://sourceware.org/gdb/wiki/ContributionChecklist#FSF_copyright_Assignment
> 
> Thank you again for submitting a patch to fix this.
> 
> Keith

Re: [PATCH 2/2] gdb: Tab complete internalvars in expressions

[Eli Zaretskii]

> Date: Wed, 28 Aug 2024 01:36:28 +0000
> From: nt8r@protonmail.com
> Cc: gdb-patches@sourceware.org
> 
> Thanks for the enthusiasm! I'm adding some tests in v2.
> 
> How do I go about copyright assignment? The documents I'm seeing seem to suggest getting in touch with a maintainer over e-mail.

I've sent the assignment form off-list.