[PATCH v3 1/5] [dap & linetable]: Add column to linetable entry

[PATCH v3 1/5] [dap & linetable]: Add column to linetable entry

[Simon Farre]

This is the first patch in a series meant to introduce the functionality
required to serve the breakpointLocations request in the DAP protocol.
It's useful for end users for other reasons, beyond that though.

For instance with this data exposed to the end user (via Python,
see later patches), a user can implement "step to next breakpoint
location on line" on their own, which maybe will be more easily accepted
than previous attempts of implementing a "next expression" command that
ended up in limbo for various reasons. Such behavior would also provide
the functionality required to serve DAP step/next requests with a
stepping granularity of "statement".

Non-DWARF debug info will just store the value 0 as the column.

I specifically made the column field 29 bits, as this means
linetable_entry will not increase in size, which seemingly is/was an
issue, reading the comment above `linetable_entry`'s definition,
though this comment is 25 years old so it may not be relevant anymore.
29 bits can represent about 500 million
(29 bits=~500 million).

This comment is 25 years old, so it might not be relevant any
longer, but I figured making it 29 bits makes it impossible to argue
against storing it (since the size doesn't change).

No changes has been made to sorting of the entries etc. They work just
like before, just that now entries store a "column" field.

v2:
Fix breaking tests

watchpoint.exp uses until for no reason really. And because it uses
until, it expects to step out of the for loop - but since we now record
"logical breakpoint locations" on a line (i.e the column metadata
provided by the compilers) "until" may not necessarily "move forward"
(unless, of course, explicitly given a line number).

By "no reason", I mean that no watchpoint is triggered during an "until"
command, so there's no reason to use in the test.

After discussion on IRC, the `until` command probably has a bug, because
it's expected to move to the "next line" and step out of a for loop,
for instance.

v3:

Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=31271
---
 gdb/buildsym-legacy.c                         |  4 ++--
 gdb/buildsym-legacy.h                         |  2 +-
 gdb/buildsym.c                                |  3 ++-
 gdb/buildsym.h                                |  4 ++--
 gdb/coffread.c                                |  4 ++--
 gdb/dbxread.c                                 |  6 ++---
 gdb/dwarf2/read.c                             | 24 +++++++++++++++----
 gdb/mdebugread.c                              |  2 +-
 gdb/symtab.h                                  |  9 +++++--
 gdb/testsuite/gdb.base/watchpoint.exp         | 16 ++++---------
 .../dw2-out-of-range-end-of-seq.exp           |  2 +-
 11 files changed, 44 insertions(+), 32 deletions(-)

diff --git a/gdb/buildsym-legacy.c b/gdb/buildsym-legacy.c
index 84bc96e843b..f267896a3b6 100644
--- a/gdb/buildsym-legacy.c
+++ b/gdb/buildsym-legacy.c
@@ -205,12 +205,12 @@ finish_block (struct symbol *symbol, struct pending_block *old_blocks,
 }
 
 void
-record_line (struct subfile *subfile, int line, unrelocated_addr pc)
+record_line (struct subfile *subfile, int line, uint col, unrelocated_addr pc)
 {
   gdb_assert (buildsym_compunit != nullptr);
   /* Assume every line entry is a statement start, that is a good place to
      put a breakpoint for that line number.  */
-  buildsym_compunit->record_line (subfile, line, pc, LEF_IS_STMT);
+  buildsym_compunit->record_line (subfile, line, col, pc, LEF_IS_STMT);
 }
 
 /* Start a new compunit_symtab for a new source file in OBJFILE.  Called, for
diff --git a/gdb/buildsym-legacy.h b/gdb/buildsym-legacy.h
index 1fb545831b8..45eacb15d30 100644
--- a/gdb/buildsym-legacy.h
+++ b/gdb/buildsym-legacy.h
@@ -76,7 +76,7 @@ extern struct context_stack *push_context (int desc, CORE_ADDR valu);
 
 extern struct context_stack pop_context ();
 
-extern void record_line (struct subfile *subfile, int line,
+extern void record_line (struct subfile *subfile, int line, uint col,
 			 unrelocated_addr pc);
 
 extern struct compunit_symtab *start_compunit_symtab (struct objfile *objfile,
diff --git a/gdb/buildsym.c b/gdb/buildsym.c
index a963219a0d2..75229d6ccda 100644
--- a/gdb/buildsym.c
+++ b/gdb/buildsym.c
@@ -623,7 +623,7 @@ buildsym_compunit::pop_subfile ()
    line vector for SUBFILE.  */
 
 void
-buildsym_compunit::record_line (struct subfile *subfile, int line,
+buildsym_compunit::record_line (struct subfile *subfile, int line, uint col,
 				unrelocated_addr pc, linetable_entry_flags flags)
 {
   m_have_line_numbers = true;
@@ -664,6 +664,7 @@ buildsym_compunit::record_line (struct subfile *subfile, int line,
   subfile->line_vector_entries.emplace_back ();
   linetable_entry &e = subfile->line_vector_entries.back ();
   e.line = line;
+  e.col = col;
   e.is_stmt = (flags & LEF_IS_STMT) != 0;
   e.set_unrelocated_pc (pc);
   e.prologue_end = (flags & LEF_PROLOGUE_END) != 0;
diff --git a/gdb/buildsym.h b/gdb/buildsym.h
index 79ece794559..a7f9c45b936 100644
--- a/gdb/buildsym.h
+++ b/gdb/buildsym.h
@@ -243,8 +243,8 @@ struct buildsym_compunit
 
   const char *pop_subfile ();
 
-  void record_line (struct subfile *subfile, int line, unrelocated_addr pc,
-		    linetable_entry_flags flags);
+  void record_line (struct subfile *subfile, int line, uint col,
+  		    unrelocated_addr pc, linetable_entry_flags flags);
 
   struct compunit_symtab *get_compunit_symtab ()
   {
diff --git a/gdb/coffread.c b/gdb/coffread.c
index 30a38035c52..77f9c11146b 100644
--- a/gdb/coffread.c
+++ b/gdb/coffread.c
@@ -1097,7 +1097,7 @@ coff_symtab_read (minimal_symbol_reader &reader,
 		 other statement-line-number.  */
 	      if (fcn_last_line == 1)
 		record_line
-		  (get_current_subfile (), fcn_first_line,
+		  (get_current_subfile (), fcn_first_line, 0,
 		   unrelocated_addr (gdbarch_addr_bits_remove (gdbarch,
 							       fcn_first_line_addr)));
 	      else
@@ -1430,7 +1430,7 @@ enter_linenos (file_ptr file_offset, int first_line,
 	{
 	  CORE_ADDR addr = lptr.l_addr.l_paddr;
 	  record_line (get_current_subfile (),
-		       first_line + L_LNNO32 (&lptr),
+		       first_line + L_LNNO32 (&lptr), 0,
 		       unrelocated_addr (gdbarch_addr_bits_remove (gdbarch,
 								   addr)));
 	}
diff --git a/gdb/dbxread.c b/gdb/dbxread.c
index 1734e8062fb..12a3d877ff9 100644
--- a/gdb/dbxread.c
+++ b/gdb/dbxread.c
@@ -2478,7 +2478,7 @@ process_one_symbol (int type, int desc, CORE_ADDR valu, const char *name,
 	      CORE_ADDR addr = last_function_start + valu;
 
 	      record_line
-		(get_current_subfile (), 0,
+		(get_current_subfile (), 0, 0,
 		 unrelocated_addr (gdbarch_addr_bits_remove (gdbarch, addr)
 				   - objfile->text_section_offset ()));
 	    }
@@ -2688,14 +2688,14 @@ process_one_symbol (int type, int desc, CORE_ADDR valu, const char *name,
 			   last_function_start : valu;
 
 	  record_line
-	    (get_current_subfile (), desc,
+	    (get_current_subfile (), desc, 0,
 	     unrelocated_addr (gdbarch_addr_bits_remove (gdbarch, addr)
 			       - objfile->text_section_offset ()));
 	  sline_found_in_function = 1;
 	}
       else
 	record_line
-	  (get_current_subfile (), desc,
+	  (get_current_subfile (), desc, 0,
 	   unrelocated_addr (gdbarch_addr_bits_remove (gdbarch, valu)
 			     - objfile->text_section_offset ()));
       break;
diff --git a/gdb/dwarf2/read.c b/gdb/dwarf2/read.c
index a50248c4d56..a4f62e6c59e 100644
--- a/gdb/dwarf2/read.c
+++ b/gdb/dwarf2/read.c
@@ -18114,6 +18114,10 @@ class lnp_state_machine
     advance_line (line_delta);
   }
 
+  void set_column (unsigned int column) noexcept {
+    m_col = column;
+  }
+
   /* Handle DW_LNS_set_file.  */
   void handle_set_file (file_name_index file);
 
@@ -18185,6 +18189,8 @@ class lnp_state_machine
   file_name_index m_file = 1;
   unsigned int m_line = 1;
 
+  unsigned int m_col = 1;
+
   /* These are initialized in the constructor.  */
 
   unrelocated_addr m_address;
@@ -18215,6 +18221,8 @@ class lnp_state_machine
      consecutive entries for the same line.  This can happen, for
      example, when discriminators are present.  PR 17276.  */
   unsigned int m_last_line = 0;
+  unsigned int m_last_col = 0;
+
   bool m_line_has_non_zero_discriminator = false;
 };
 
@@ -18318,6 +18326,7 @@ lnp_state_machine::handle_const_add_pc ()
 static int
 dwarf_record_line_p (struct dwarf2_cu *cu,
 		     unsigned int line, unsigned int last_line,
+		     uint col, uint last_col,
 		     int line_has_non_zero_discriminator,
 		     struct subfile *last_subfile)
 {
@@ -18325,6 +18334,8 @@ dwarf_record_line_p (struct dwarf2_cu *cu,
     return 1;
   if (line != last_line)
     return 1;
+  else if(line == last_line && col != last_col)
+    return 1;
   /* Same line for the same file that we've seen already.
      As a last check, for pr 17276, only record the line if the line
      has never had a non-zero discriminator.  */
@@ -18338,7 +18349,7 @@ dwarf_record_line_p (struct dwarf2_cu *cu,
 
 static void
 dwarf_record_line_1 (struct gdbarch *gdbarch, struct subfile *subfile,
-		     unsigned int line, unrelocated_addr address,
+		     unsigned int line, uint col, unrelocated_addr address,
 		     linetable_entry_flags flags,
 		     struct dwarf2_cu *cu)
 {
@@ -18355,7 +18366,7 @@ dwarf_record_line_1 (struct gdbarch *gdbarch, struct subfile *subfile,
     }
 
   if (cu != nullptr)
-    cu->get_builder ()->record_line (subfile, line, addr, flags);
+    cu->get_builder ()->record_line (subfile, line, col, addr, flags);
 }
 
 /* Subroutine of dwarf_decode_lines_1 to simplify it.
@@ -18378,7 +18389,7 @@ dwarf_finish_line (struct gdbarch *gdbarch, struct subfile *subfile,
 		  paddress (gdbarch, (CORE_ADDR) address));
     }
 
-  dwarf_record_line_1 (gdbarch, subfile, 0, address, LEF_IS_STMT, cu);
+  dwarf_record_line_1 (gdbarch, subfile, 0, 0, address, LEF_IS_STMT, cu);
 }
 
 void
@@ -18444,17 +18455,19 @@ lnp_state_machine::record_line (bool end_sequence)
 	    lte_flags |= LEF_IS_STMT;
 
 	  if (dwarf_record_line_p (m_cu, m_line, m_last_line,
+	  			   m_col, m_last_col,
 				   m_line_has_non_zero_discriminator,
 				   m_last_subfile))
 	    {
 	      buildsym_compunit *builder = m_cu->get_builder ();
 	      dwarf_record_line_1 (m_gdbarch,
 				   builder->get_current_subfile (),
-				   m_line, m_address, lte_flags,
+				   m_line, m_col, m_address, lte_flags,
 				   m_currently_recording_lines ? m_cu : nullptr);
 	    }
 	  m_last_subfile = m_cu->get_builder ()->get_current_subfile ();
 	  m_last_line = m_line;
+	  m_last_col = m_col;
 	}
     }
 
@@ -18670,7 +18683,8 @@ dwarf_decode_lines_1 (struct line_header *lh, struct dwarf2_cu *cu,
 	      }
 	      break;
 	    case DW_LNS_set_column:
-	      (void) read_unsigned_leb128 (abfd, line_ptr, &bytes_read);
+	      state_machine.set_column (
+		read_unsigned_leb128 (abfd, line_ptr, &bytes_read));
 	      line_ptr += bytes_read;
 	      break;
 	    case DW_LNS_negate_stmt:
diff --git a/gdb/mdebugread.c b/gdb/mdebugread.c
index 73df0ed8165..b5cbe4548a7 100644
--- a/gdb/mdebugread.c
+++ b/gdb/mdebugread.c
@@ -4008,7 +4008,7 @@ mdebug_expand_psymtab (legacy_psymtab *pst, struct objfile *objfile)
 		{
 		  /* Handle encoded stab line number.  */
 		  record_line
-		    (get_current_subfile (), sh.index,
+		    (get_current_subfile (), sh.index, 0,
 		     unrelocated_addr (gdbarch_addr_bits_remove (gdbarch,
 								 valu)));
 		}
diff --git a/gdb/symtab.h b/gdb/symtab.h
index eecd999b7e6..38d08fe8599 100644
--- a/gdb/symtab.h
+++ b/gdb/symtab.h
@@ -1616,14 +1616,19 @@ struct linetable_entry
     return m_pc < other.m_pc;
   }
 
-  /* Two entries are equal if they have the same line and PC.  The
+  /* Two entries are equal if they have the same line, PC and col.  The
      other members are ignored.  */
   bool operator== (const linetable_entry &other) const
-  { return line == other.line && m_pc == other.m_pc; }
+  { return line == other.line && m_pc == other.m_pc && col == other.col; }
 
   /* The line number for this entry.  */
   int line;
 
+  /* The column number for entry.  Assume a maximum of circa 500M columns,
+     a similar assumption is made for line after all.  Keeps an identical
+     size for linetable_entry (16 bytes).  */
+  unsigned int col : 29;
+
   /* True if this PC is a good location to place a breakpoint for LINE.  */
   bool is_stmt : 1;
 
diff --git a/gdb/testsuite/gdb.base/watchpoint.exp b/gdb/testsuite/gdb.base/watchpoint.exp
index 2449663ef8b..83fc42c0f6c 100644
--- a/gdb/testsuite/gdb.base/watchpoint.exp
+++ b/gdb/testsuite/gdb.base/watchpoint.exp
@@ -327,19 +327,11 @@ proc test_stepping {} {
 	# stepping through the loop once, and the debugger should not
 	# stop for any of the remaining iterations.
 
-	gdb_test "until" "ival1 = count.*" "until to ival1 assignment"
-	gdb_test "until" "ival3 = count.*" "until to ival3 assignment"
-	set test "until out of loop"
-	gdb_test_multiple "until" $test {
-	    -re "(for \\(count = 0|\}).*$gdb_prompt $" {
-		gdb_test "until" "ival1 = count; /. Outside loop ./" $test
-	    }
-	    -re "ival1 = count; /. Outside loop ./.*$gdb_prompt $" {
-		pass $test
-	    }
-	}
+	gdb_test "tbreak watchpoint.c:185" \
+		".*watchpoint.c:185.*" \
+		"set temporarybreakpoint to before assignment of ival2"
 
-	gdb_test "step" "ival2 = count.*" "step to ival2 assignment"
+	gdb_test "continue" "ival2 = count.*" "cont to ival2 assignment"
 
 	# Check that the watchpoint is triggered during a step.
 	gdb_test "step" \
diff --git a/gdb/testsuite/gdb.dwarf2/dw2-out-of-range-end-of-seq.exp b/gdb/testsuite/gdb.dwarf2/dw2-out-of-range-end-of-seq.exp
index 060b5b6dbfa..33c417576e6 100644
--- a/gdb/testsuite/gdb.dwarf2/dw2-out-of-range-end-of-seq.exp
+++ b/gdb/testsuite/gdb.dwarf2/dw2-out-of-range-end-of-seq.exp
@@ -84,7 +84,7 @@ if ![runto_main] {
 }
 
 set test "END with address 1 eliminated"
-gdb_test_multiple "maint info line-table \\b$srcfile$" $test {
+gdb_test_multiple "maint info line-table $srcfile" $test {
     -re -wrap "END *0x0*1 *$hex *Y *\r\n.*" {
 	fail $gdb_test_name
     }
-- 
2.43.0

[PATCH v3 2/5] [dap & linetable]: Add column to maint info linetable output

[Simon Farre]

This adds "COL" to the output of the maintenance command "info
line-table"

v2.
maint info line-table: Make display of column optional

Toggle the display (on) of column meta data by adding `-show-col`
to the command. This command is used way too much in testing to change
at this point making any feature that touches this almost impossible to
move forward due to test maintenance hell.

v3:

Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=31271
---
 gdb/symmisc.c | 102 +++++++++++++++++++++++++++++++++++++-------------
 1 file changed, 76 insertions(+), 26 deletions(-)

diff --git a/gdb/symmisc.c b/gdb/symmisc.c
index 5dfb9fca8f0..b7dccb3893e 100644
--- a/gdb/symmisc.c
+++ b/gdb/symmisc.c
@@ -937,17 +937,67 @@ block_depth (const struct block *block)
     }
   return i;
 }
-\f
+
+static void
+output_lt_row (ui_out *uiout, int i, const linetable_entry &item,
+	       const objfile *obj)
+{
+  ui_out_emit_tuple tuple_emitter (uiout, nullptr);
+  uiout->field_signed ("index", i);
+  if (item.line > 0)
+    {
+      uiout->field_signed ("line", item.line);
+    }
+  else
+    {
+      uiout->field_string ("line", _ ("END"));
+    }
+  uiout->field_core_addr ("rel-address", obj->arch (), item.pc (obj));
+  uiout->field_core_addr ("unrel-address", obj->arch (),
+			  CORE_ADDR (item.unrelocated_pc ()));
+  uiout->field_string ("is-stmt", item.is_stmt ? "Y" : "");
+  uiout->field_string ("prologue-end", item.prologue_end ? "Y" : "");
+  uiout->field_string ("epilogue-begin", item.epilogue_begin ? "Y" : "");
+  uiout->text ("\n");
+}
+
+static auto
+output_lt_row_with_col (ui_out *uiout, int i, const linetable_entry &item,
+			const objfile *obj)
+{
+  ui_out_emit_tuple tuple_emitter (uiout, nullptr);
+  uiout->field_signed ("index", i);
+  if (item.line > 0)
+    {
+      uiout->field_signed ("line", item.line);
+      uiout->field_signed ("col", item.col);
+    }
+  else
+    {
+      uiout->field_string ("line", _ ("END"));
+      uiout->field_string ("col", _ ("END"));
+    }
+  uiout->field_core_addr ("rel-address", obj->arch (), item.pc (obj));
+  uiout->field_core_addr ("unrel-address", obj->arch (),
+			  CORE_ADDR (item.unrelocated_pc ()));
+  uiout->field_string ("is-stmt", item.is_stmt ? "Y" : "");
+  uiout->field_string ("prologue-end", item.prologue_end ? "Y" : "");
+  uiout->field_string ("epilogue-begin", item.epilogue_begin ? "Y" : "");
+  uiout->text ("\n");
+}
 
 /* Used by MAINTENANCE_INFO_LINE_TABLES to print the information about a
    single line table.  */
 
 static int
-maintenance_print_one_line_table (struct symtab *symtab, void *data)
+maintenance_print_one_line_table (struct symtab *symtab, void *data,
+				  bool display_col)
 {
   const struct linetable *linetable;
   struct objfile *objfile;
 
+  const auto columns = display_col ? 8 : 7;
+
   objfile = symtab->compunit ()->objfile ();
   gdb_printf (_("objfile: %ps ((struct objfile *) %s)\n"),
 	      styled_string (file_name_style.style (),
@@ -973,9 +1023,11 @@ maintenance_print_one_line_table (struct symtab *symtab, void *data)
       /* Leave space for 6 digits of index and line number.  After that the
 	 tables will just not format as well.  */
       struct ui_out *uiout = current_uiout;
-      ui_out_emit_table table_emitter (uiout, 7, -1, "line-table");
+      ui_out_emit_table table_emitter (uiout, columns, -1, "line-table");
       uiout->table_header (6, ui_left, "index", _("INDEX"));
       uiout->table_header (6, ui_left, "line", _("LINE"));
+      if (display_col)
+	uiout->table_header (6, ui_left, "col", _("COL"));
       uiout->table_header (18, ui_left, "rel-address", _("REL-ADDRESS"));
       uiout->table_header (18, ui_left, "unrel-address", _("UNREL-ADDRESS"));
       uiout->table_header (7, ui_left, "is-stmt", _("IS-STMT"));
@@ -983,26 +1035,12 @@ maintenance_print_one_line_table (struct symtab *symtab, void *data)
       uiout->table_header (14, ui_left, "epilogue-begin", _("EPILOGUE-BEGIN"));
       uiout->table_body ();
 
-      for (int i = 0; i < linetable->nitems; ++i)
-	{
-	  const linetable_entry *item;
-
-	  item = &linetable->item [i];
-	  ui_out_emit_tuple tuple_emitter (uiout, nullptr);
-	  uiout->field_signed ("index", i);
-	  if (item->line > 0)
-	    uiout->field_signed ("line", item->line);
-	  else
-	    uiout->field_string ("line", _("END"));
-	  uiout->field_core_addr ("rel-address", objfile->arch (),
-				  item->pc (objfile));
-	  uiout->field_core_addr ("unrel-address", objfile->arch (),
-				  CORE_ADDR (item->unrelocated_pc ()));
-	  uiout->field_string ("is-stmt", item->is_stmt ? "Y" : "");
-	  uiout->field_string ("prologue-end", item->prologue_end ? "Y" : "");
-	  uiout->field_string ("epilogue-begin", item->epilogue_begin ? "Y" : "");
-	  uiout->text ("\n");
-	}
+      if (display_col)
+	for (int i = 0; i < linetable->nitems; ++i)
+	  output_lt_row_with_col (uiout, i, linetable->item[i], objfile);
+      else
+	for (int i = 0; i < linetable->nitems; ++i)
+	  output_lt_row (uiout, i, linetable->item[i], objfile);
     }
 
   return 0;
@@ -1011,9 +1049,20 @@ maintenance_print_one_line_table (struct symtab *symtab, void *data)
 /* Implement the 'maint info line-table' command.  */
 
 static void
-maintenance_info_line_tables (const char *regexp, int from_tty)
+maintenance_info_line_tables (const char* args, int from_tty)
 {
   dont_repeat ();
+  const char* regexp = nullptr;
+  bool show_column = false;
+  gdb_argv argv (args);
+
+  for (auto i = 0; argv != nullptr && argv[i] != nullptr; ++i)
+    {
+      if (strcmp (argv[i], "-show-col") == 0)
+	show_column = true;
+      else
+	regexp = argv[i];
+    }
 
   if (regexp != NULL)
     re_comp (regexp);
@@ -1030,7 +1079,7 @@ maintenance_info_line_tables (const char *regexp, int from_tty)
 		if (regexp == NULL
 		    || re_exec (symtab_to_filename_for_display (symtab)))
 		  {
-		    maintenance_print_one_line_table (symtab, NULL);
+		    maintenance_print_one_line_table (symtab, NULL, show_column);
 		    gdb_printf ("\n");
 		  }
 	      }
@@ -1080,7 +1129,8 @@ With an argument REGEXP, list the symbol tables with matching names."),
   add_cmd ("line-table", class_maintenance, maintenance_info_line_tables, _("\
 List the contents of all line tables, from all symbol tables.\n\
 With an argument REGEXP, list just the line tables for the symbol\n\
-tables with matching names."),
+tables with matching names.\n\
+With the argument -show-col this command also displays column information."),
 	   &maintenanceinfolist);
 
   add_cmd ("check-symtabs", class_maintenance, maintenance_check_symtabs,
-- 
2.43.0

[PATCH v3 3/5] [dap & linetable]: Change gdb.LineTableEntry & Add gdb.lookup_linetable

[Simon Farre]

This patch changes the Python type `LineTableEntry` by adding the
`column` accessor which returns the newly added `column` value from the
internal linetable_entry.

It also exposes a new "module function": gdb.lookup_linetable which
searches all symbtabs using a string; comparing it with the symtabs
'fullname' and returning the linetable of the (first) symtab that
matches.

To search for a linetable from python:

lt = gdb.lookup_linetable("main.cpp")

And like before (as per current documentation), one can use that like so:

for lte in lt:
  print(f"line:col={lte.line}:{lte.column}; pc = {hex(lte.pc)}")

Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=31271
---
 gdb/python/py-linetable.c    | 35 ++++++++++++++++++++++++-----------
 gdb/python/py-symtab.c       | 33 +++++++++++++++++++++++++++++++++
 gdb/python/python-internal.h |  2 ++
 gdb/python/python.c          |  2 ++
 gdb/symtab.c                 | 19 ++++++++++++++++++-
 gdb/symtab.h                 |  3 +++
 6 files changed, 82 insertions(+), 12 deletions(-)

diff --git a/gdb/python/py-linetable.c b/gdb/python/py-linetable.c
index 788a6e1e24b..ba96c663027 100644
--- a/gdb/python/py-linetable.c
+++ b/gdb/python/py-linetable.c
@@ -24,6 +24,8 @@ struct linetable_entry_object {
   PyObject_HEAD
   /* The line table source line.  */
   int line;
+  /* The line table source column.  */
+  int col;
   /* The pc associated with the source line.  */
   CORE_ADDR pc;
 };
@@ -99,7 +101,7 @@ symtab_to_linetable_object (PyObject *symtab)
    and an address.  */
 
 static PyObject *
-build_linetable_entry (int line, CORE_ADDR address)
+build_linetable_entry (int line, int col, CORE_ADDR address)
 {
   linetable_entry_object *obj;
 
@@ -109,6 +111,7 @@ build_linetable_entry (int line, CORE_ADDR address)
     {
       obj->line = line;
       obj->pc = address;
+      obj->col = col;
     }
 
   return (PyObject *) obj;
@@ -121,7 +124,7 @@ build_linetable_entry (int line, CORE_ADDR address)
    address.  */
 
 static PyObject *
-build_line_table_tuple_from_pcs (int line, const std::vector<CORE_ADDR> &pcs)
+build_line_table_tuple_from_pcs (int line, const symtab& sym, const std::vector<linetable_entry> &pcs)
 {
   int i;
 
@@ -133,10 +136,11 @@ build_line_table_tuple_from_pcs (int line, const std::vector<CORE_ADDR> &pcs)
   if (tuple == NULL)
     return NULL;
 
+  const auto of = sym.compunit ()->objfile ();
   for (i = 0; i < pcs.size (); ++i)
     {
-      CORE_ADDR pc = pcs[i];
-      gdbpy_ref<> obj (build_linetable_entry (line, pc));
+      const auto& lte = pcs[i];
+      gdbpy_ref<> obj (build_linetable_entry (line, lte.col, lte.pc (of)));
 
       if (obj == NULL)
 	return NULL;
@@ -152,12 +156,10 @@ build_line_table_tuple_from_pcs (int line, const std::vector<CORE_ADDR> &pcs)
    in the line table.  */
 
 static PyObject *
-ltpy_get_pcs_for_line (PyObject *self, PyObject *args)
+ltpy_get_ltes_for_line (PyObject *self, PyObject *args)
 {
   struct symtab *symtab;
   gdb_py_longest py_line;
-  const linetable_entry *best_entry = nullptr;
-  std::vector<CORE_ADDR> pcs;
 
   LTPY_REQUIRE_VALID (self, symtab);
 
@@ -166,14 +168,15 @@ ltpy_get_pcs_for_line (PyObject *self, PyObject *args)
 
   try
     {
-      pcs = find_pcs_for_symtab_line (symtab, py_line, &best_entry);
+      const auto ltes = find_ltes_for_symtab_line (*symtab, py_line);
+      return build_line_table_tuple_from_pcs (py_line, *symtab, ltes);
     }
   catch (const gdb_exception &except)
     {
       GDB_PY_HANDLE_EXCEPTION (except);
     }
 
-  return build_line_table_tuple_from_pcs (py_line, pcs);
+  Py_RETURN_NONE;
 }
 
 /* Implementation of gdb.LineTable.has_line (self, line) -> Boolean.
@@ -327,6 +330,14 @@ ltpy_entry_get_line (PyObject *self, void *closure)
   return gdb_py_object_from_longest (obj->line).release ();
 }
 
+static PyObject *
+ltpy_entry_get_column (PyObject *self, void *closure)
+{
+  linetable_entry_object *obj = (linetable_entry_object *) self;
+
+  return gdb_py_object_from_longest (obj->col).release ();
+}
+
 /* Implementation of gdb.LineTableEntry.pc (self) -> Long.  Returns a
    a long integer associated with the PC of the line table entry.  */
 
@@ -422,7 +433,7 @@ ltpy_iternext (PyObject *self)
     }
 
   struct objfile *objfile = symtab->compunit ()->objfile ();
-  obj = build_linetable_entry (item->line, item->pc (objfile));
+  obj = build_linetable_entry (item->line, item->col, item->pc (objfile));
   iter_obj->current_index++;
 
   return obj;
@@ -451,7 +462,7 @@ GDBPY_INITIALIZE_FILE (gdbpy_initialize_linetable);
 \f
 
 static PyMethodDef linetable_object_methods[] = {
-  { "line", ltpy_get_pcs_for_line, METH_VARARGS,
+  { "line", ltpy_get_ltes_for_line, METH_VARARGS,
     "line (lineno) -> Tuple\n\
 Return executable locations for a given source line." },
   { "has_line", ltpy_has_line, METH_VARARGS,
@@ -548,6 +559,8 @@ PyTypeObject ltpy_iterator_object_type = {
 static gdb_PyGetSetDef linetable_entry_object_getset[] = {
   { "line", ltpy_entry_get_line, NULL,
     "The line number in the source file.", NULL },
+  { "column", ltpy_entry_get_column, NULL,
+    "The column number in the source file", NULL },
   { "pc", ltpy_entry_get_pc, NULL,
     "The memory address for this line number.", NULL },
   { NULL }  /* Sentinel */
diff --git a/gdb/python/py-symtab.c b/gdb/python/py-symtab.c
index b28f6ca8b48..e23c02aa0e4 100644
--- a/gdb/python/py-symtab.c
+++ b/gdb/python/py-symtab.c
@@ -475,6 +475,39 @@ symtab_to_symtab_object (struct symtab *symtab)
   return (PyObject *) symtab_obj;
 }
 
+/* Search for a symtab whose fullname contains FILENAME and return it's
+   linetable.  */
+PyObject *
+gdbpy_lookup_linetable_by_filename (PyObject *, PyObject *args, PyObject *kw)
+{
+  const char *arg = nullptr;
+  static const char *keywords[] = { "filename", nullptr };
+
+  if (!gdb_PyArg_ParseTupleAndKeywords (args, kw, "s", keywords, &arg))
+    {
+      return nullptr;
+    }
+
+  for (objfile *objfile : current_program_space->objfiles ())
+    for (compunit_symtab *cu : objfile->compunits ())
+      for (symtab *s : cu->filetabs ())
+	{
+	  const auto name = symtab_to_fullname (s);
+
+	  if (strstr (name, arg))
+	    {
+	      auto symtab_pyobj = symtab_to_symtab_object (s);
+	      if (!symtab_pyobj)
+		Py_RETURN_NONE;
+	      auto lt_obj = symtab_to_linetable_object (symtab_pyobj);
+	      if (!lt_obj)
+		Py_RETURN_NONE;
+	      return lt_obj;
+	    }
+	}
+  Py_RETURN_NONE;
+}
+
 /* Create a new symtab and line (gdb.Symtab_and_line) object
    that encapsulates the symtab_and_line structure from GDB.  */
 PyObject *
diff --git a/gdb/python/python-internal.h b/gdb/python/python-internal.h
index 7c05007cbab..dcfc03e3beb 100644
--- a/gdb/python/python-internal.h
+++ b/gdb/python/python-internal.h
@@ -408,6 +408,8 @@ PyObject *gdbpy_convenience_variable (PyObject *self, PyObject *args);
 PyObject *gdbpy_set_convenience_variable (PyObject *self, PyObject *args);
 PyObject *gdbpy_breakpoints (PyObject *, PyObject *);
 PyObject *gdbpy_frame_stop_reason_string (PyObject *, PyObject *);
+PyObject *gdbpy_lookup_linetable_by_filename (PyObject *, PyObject* args,
+					      PyObject *kw);
 PyObject *gdbpy_lookup_symbol (PyObject *self, PyObject *args, PyObject *kw);
 PyObject *gdbpy_lookup_global_symbol (PyObject *self, PyObject *args,
 				      PyObject *kw);
diff --git a/gdb/python/python.c b/gdb/python/python.c
index 1d406392bd3..3d7c70fed09 100644
--- a/gdb/python/python.c
+++ b/gdb/python/python.c
@@ -2573,6 +2573,8 @@ gdbpy_initialized (const struct extension_language_defn *extlang)
 
 PyMethodDef python_GdbMethods[] =
 {
+  { "lookup_linetable", (PyCFunction) gdbpy_lookup_linetable_by_filename,
+    METH_VARARGS | METH_KEYWORDS },
   { "history", gdbpy_history, METH_VARARGS,
     "Get a value from history" },
   { "add_history", gdbpy_add_history, METH_VARARGS,
diff --git a/gdb/symtab.c b/gdb/symtab.c
index 7c0a69108d4..e206ba9cada 100644
--- a/gdb/symtab.c
+++ b/gdb/symtab.c
@@ -3519,7 +3519,24 @@ find_pcs_for_symtab_line (struct symtab *symtab, int line,
   return result;
 }
 
-\f
+/**  Return line table entries for SYMTAB at LINE.  */
+
+std::vector<linetable_entry>
+find_ltes_for_symtab_line (const symtab &symtab, int line)
+{
+  auto lte = symtab.linetable ();
+  if (lte == nullptr)
+    return {};
+
+  std::vector<linetable_entry> result;
+  for (auto i = 0; i < lte->nitems; ++i)
+    {
+      if (lte->item[i].line == line)
+	result.push_back (lte->item[i]);
+    }
+  return result;
+}
+
 /* Set the PC value for a given source file and line number and return true.
    Returns false for invalid line number (and sets the PC to 0).
    The source file is specified with a struct symtab.  */
diff --git a/gdb/symtab.h b/gdb/symtab.h
index 38d08fe8599..9098822c90b 100644
--- a/gdb/symtab.h
+++ b/gdb/symtab.h
@@ -2772,6 +2772,9 @@ void iterate_over_symtabs (const char *name,
 std::vector<CORE_ADDR> find_pcs_for_symtab_line
     (struct symtab *symtab, int line, const linetable_entry **best_entry);
 
+std::vector<linetable_entry> find_ltes_for_symtab_line
+    (const symtab &symtab, int line);
+
 /* Prototype for callbacks for LA_ITERATE_OVER_SYMBOLS.  The callback
    is called once per matching symbol SYM.  The callback should return
    true to indicate that LA_ITERATE_OVER_SYMBOLS should continue
-- 
2.43.0

[PATCH v3 4/5] [dap & linetable]: Add breakpointLocations request

[Simon Farre]

This adds the breakpointLocations request.

Testing might strictly not be possible for this feature. Users are going
to have wildly varying compilers and compiler versions that outputs
wildly different debug symbol information. So unless we can restrain the
tests to some specific toolchain that I can handroll the tests for, this
will not be useful (and handrolling for a specific compiler defeats the
purpose - as one can just use the eyes then to verify).

v2:
Fixed type check issue

Forgot the obligatory **args at the end of breakpoint_locations

v3:

Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=31271
---
 gdb/python/lib/gdb/dap/breakpoint.py | 27 +++++++++++++++++++++++++++
 1 file changed, 27 insertions(+)

diff --git a/gdb/python/lib/gdb/dap/breakpoint.py b/gdb/python/lib/gdb/dap/breakpoint.py
index 9cbd7ae0c47..70c93b0a03d 100644
--- a/gdb/python/lib/gdb/dap/breakpoint.py
+++ b/gdb/python/lib/gdb/dap/breakpoint.py
@@ -437,3 +437,30 @@ def set_exception_breakpoints(
     return {
         "breakpoints": _set_exception_catchpoints(options),
     }
+
+
+@in_gdb_thread
+def _get_logical_bp_locations(name: str, line: int):
+    linetable = gdb.lookup_linetable(name)
+    if linetable is not None:
+        return [
+            {"line": lte.line, "column": lte.column} for lte in linetable.line(line)
+        ]
+    else:
+        return []
+
+
+@request("breakpointLocations")
+@capability("supportsBreakpointLocationsRequest")
+def breakpoint_locations(
+    *,
+    # This is a Source but we don't type-check it.
+    source,
+    line: int,
+    column: Optional[int],
+    endLine: Optional[int],
+    endColumn: Optional[int],
+    **args,
+):
+    # for now we don't support column, endLine, endColumn. We just return per line
+    return {"breakpoints": _get_logical_bp_locations(source["path"], line)}
-- 
2.43.0

[PATCH v3 5/5] [dap & linetable]: Added docs

[Simon Farre]

Added docs for the new `gdb.lookup_linetable` as well as the new
accessor method `LineTableEntry.column`.

Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=31271
---
 gdb/doc/python.texi | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)

diff --git a/gdb/doc/python.texi b/gdb/doc/python.texi
index 71431878dd3..3206aef3b4f 100644
--- a/gdb/doc/python.texi
+++ b/gdb/doc/python.texi
@@ -6422,6 +6422,18 @@ mapping of source lines to their executable locations in memory.  To
 acquire the line table information for a particular symbol table, use
 the @code{linetable} function (@pxref{Symbol Tables In Python}).
 
+Python code can request a symtab's linetable by calling
+@code{gdb.lookup_linetable} passing in the filename.  The parameter
+@var{filename} can be a substring of the symtab's @code{fullname}.
+If multiple symtabs' fullnames share that substring, the first one
+found is returned.
+
+@defun gdb.lookup_linetable (filename)
+Return the @code{gdb.LineTable} of a symtab whose fullname matches
+@var{filename}.  If no symtab could be found using @var{filename}
+as a substring this function returns @code{None}.
+@end defun
+
 A @code{gdb.LineTable} is iterable.  The iterator returns
 @code{LineTableEntry} objects that correspond to the source line and
 address for each line table entry.  @code{LineTableEntry} objects have
@@ -6433,6 +6445,12 @@ corresponds to the actual line of source.  This attribute is not
 writable.
 @end defvar
 
+@defvar LineTableEntry.column
+The source column number for this line table entry.  This number
+corresponds to the actual column of source.  This attribute is not
+writable.
+@end defvar
+
 @defvar LineTableEntry.pc
 The address that is associated with the line table entry where the
 executable code for that source line resides in memory.  This
-- 
2.43.0

Re: [PATCH v3 1/5] [dap & linetable]: Add column to linetable entry

[Tom Tromey]

>>>>> "Simon" == Simon Farre <simon.farre.cx@gmail.com> writes:

Thanks for the patch.

Simon>  void
Simon> -record_line (struct subfile *subfile, int line, unrelocated_addr pc)
Simon> +record_line (struct subfile *subfile, int line, uint col, unrelocated_addr pc)

"uint" isn't really a type used in gdb.  I think "unsigned" would be
fine.

 
Simon> +  void set_column (unsigned int column) noexcept {
Simon> +    m_col = column;
Simon> +  }

gdb style puts the opening brace on the next line.

Simon> +  else if(line == last_line && col != last_col)
Simon> +    return 1;

Space before paren.

Simon> diff --git a/gdb/mdebugread.c b/gdb/mdebugread.c
Simon> index 73df0ed8165..b5cbe4548a7 100644
Simon> --- a/gdb/mdebugread.c
Simon> +++ b/gdb/mdebugread.c
Simon> @@ -4008,7 +4008,7 @@ mdebug_expand_psymtab (legacy_psymtab *pst, struct objfile *objfile)
Simon>  		{
Simon>  		  /* Handle encoded stab line number.  */
Simon>  		  record_line
Simon> -		    (get_current_subfile (), sh.index,
Simon> +		    (get_current_subfile (), sh.index, 0,

Unfortunately there's also mdebugread.c:add_line.
I suspect this code is already a bit wrong because new_linetable doesn't
zero out the memory.

I think there are other such spots needing an update as well, for
instance jit.c.

Simon> -  /* Two entries are equal if they have the same line and PC.  The
Simon> +  /* Two entries are equal if they have the same line, PC and col.  The
Simon>       other members are ignored.  */
Simon>    bool operator== (const linetable_entry &other) const
Simon> -  { return line == other.line && m_pc == other.m_pc; }
Simon> +  { return line == other.line && m_pc == other.m_pc && col == other.col; }

Does operator< also need a change?

Tom

Re: [PATCH v3 4/5] [dap & linetable]: Add breakpointLocations request

[Tom Tromey]

>>>>> "Simon" == Simon Farre <simon.farre.cx@gmail.com> writes:

Simon> Testing might strictly not be possible for this feature. Users are going
Simon> to have wildly varying compilers and compiler versions that outputs
Simon> wildly different debug symbol information. So unless we can restrain the
Simon> tests to some specific toolchain that I can handroll the tests for, this
Simon> will not be useful (and handrolling for a specific compiler defeats the
Simon> purpose - as one can just use the eyes then to verify).

It may be possible to write a custom test case using the "DWARF
assembler".

Simon> +@request("breakpointLocations")
Simon> +@capability("supportsBreakpointLocationsRequest")

This already exists in locations.py.

I'll send a patch to make sure that double-registration is impossible.

Tom

Re: [PATCH v3 2/5] [dap & linetable]: Add column to maint info linetable output

[Tom Tromey]

>>>>> "Simon" == Simon Farre <simon.farre.cx@gmail.com> writes:

Simon> Toggle the display (on) of column meta data by adding `-show-col`
Simon> to the command. This command is used way too much in testing to change
Simon> at this point making any feature that touches this almost impossible to
Simon> move forward due to test maintenance hell.

When I looked it seemed like there's only a handful of spots to modify;
and since it's a "maint" command I tend to think it would be better to
just unconditionally emit the column.

Tom

Re: [PATCH v3 3/5] [dap & linetable]: Change gdb.LineTableEntry & Add gdb.lookup_linetable

[Tom Tromey]

>>>>> "Simon" == Simon Farre <simon.farre.cx@gmail.com> writes:

Thanks for the patch.

Simon>  static PyObject *
Simon> -build_line_table_tuple_from_pcs (int line, const std::vector<CORE_ADDR> &pcs)
Simon> +build_line_table_tuple_from_pcs (int line, const symtab& sym, const std::vector<linetable_entry> &pcs)

gdb uses a line length around 80, so this should be split.
Also the "&" attaches to the "sym", not "symtab" -- this is also true
for "*" in gdb, there's a few instances of this around.

Simon> +/* Search for a symtab whose fullname contains FILENAME and return it's
Simon> +   linetable.  */
Simon> +PyObject *
Simon> +gdbpy_lookup_linetable_by_filename (PyObject *, PyObject *args, PyObject *kw)
Simon> +{
Simon> +  const char *arg = nullptr;
Simon> +  static const char *keywords[] = { "filename", nullptr };
Simon> +
Simon> +  if (!gdb_PyArg_ParseTupleAndKeywords (args, kw, "s", keywords, &arg))
Simon> +    {
Simon> +      return nullptr;
Simon> +    }

Don't need the braces here.

Simon> +
Simon> +  for (objfile *objfile : current_program_space->objfiles ())
Simon> +    for (compunit_symtab *cu : objfile->compunits ())
Simon> +      for (symtab *s : cu->filetabs ())
Simon> +	{
Simon> +	  const auto name = symtab_to_fullname (s);
Simon> +
Simon> +	  if (strstr (name, arg))

This doesn't seem like a great approach to me.
Substring matching isn't how this kind of thing works elsewhere in gdb.

Also this looping approach means that the symtab must somehow have
already been expanded.

Right now, the breakpointLocations request works by calling MI's
-symbol-list-lines.  This calls lookup_symtab, which in the end will try
expanding CUs to find the desired symtab.

Tom