Mirror of the gdb-patches mailing list
 help / color / mirror / Atom feed
From: "Tom de Vries (Code Review)" <gerrit@gnutoolchain-gerrit.osci.io>
To: Luis Machado <luis.machado@linaro.org>,
	Andrew Burgess <andrew.burgess@embecosm.com>,
	gdb-patches@sourceware.org
Subject: [review v3] [gdb/symtab] Prefer var def over decl
Date: Wed, 23 Oct 2019 16:27:00 -0000	[thread overview]
Message-ID: <20191023162717.844292192E@gnutoolchain-gerrit.osci.io> (raw)
In-Reply-To: <gerrit.1571043046000.Id92326cb8ef9903b121ef9e320658eb565d0f5a9@gnutoolchain-gerrit.osci.io>

Change URL: https://gnutoolchain-gerrit.osci.io/r/c/binutils-gdb/+/29
......................................................................

[gdb/symtab] Prefer var def over decl

Consider the DWARF as generated by gcc with the tentative patch to fix gcc
PR91507 - "wrong debug for completed array with previous incomplete
declaration":
...
 <1><f4>: Abbrev Number: 2 (DW_TAG_array_type)
    <f5>   DW_AT_type        : <0xff>
    <f9>   DW_AT_sibling     : <0xff>
 <2><fd>: Abbrev Number: 3 (DW_TAG_subrange_type)
 <2><fe>: Abbrev Number: 0
 <1><ff>: Abbrev Number: 4 (DW_TAG_pointer_type)
    <100>   DW_AT_byte_size   : 8
    <101>   DW_AT_type        : <0x105>
 <1><105>: Abbrev Number: 5 (DW_TAG_base_type)
    <106>   DW_AT_byte_size   : 1
    <107>   DW_AT_encoding    : 6       (signed char)
    <108>   DW_AT_name        : (indirect string, offset: 0x19f): char
 <1><10c>: Abbrev Number: 6 (DW_TAG_variable)
    <10d>   DW_AT_name        : zzz
    <111>   DW_AT_decl_file   : 1
    <112>   DW_AT_decl_line   : 1
    <113>   DW_AT_decl_column : 14
    <114>   DW_AT_type        : <0xf4>
    <118>   DW_AT_external    : 1
    <118>   DW_AT_declaration : 1
 <1><118>: Abbrev Number: 2 (DW_TAG_array_type)
    <119>   DW_AT_type        : <0xff>
    <11d>   DW_AT_sibling     : <0x128>
 <1><12f>: Abbrev Number: 8 (DW_TAG_variable)
    <130>   DW_AT_specification: <0x10c>
    <134>   DW_AT_decl_line   : 2
    <135>   DW_AT_decl_column : 7
    <136>   DW_AT_type        : <0x118>
    <13a>   DW_AT_location    : 9 byte block: 3 30 10 60 0 0 0 0 0      (DW_OP_addr: 601030)
...

The DWARF will result in two entries in the symbol table, a decl with type
char *[] and a def with type char*[2].

When trying to print the value of zzz:
...
$ gdb a.spec.out -batch -ex "p zzz"
...
the decl (rather than the def) will be found in the symbol table, which is
missing the location information, and consequently we get:
...
$1 = 0x601030 <zzz>
...

[ There is a fallback mechanism that finds the address of the variable in the
minimal symbol table, but that's not used here, because the type of the decl
does not specify a size.  We could use the symbol size here to get the size
of the type, but that's currently not done: PR exp/24989.  Still, fixing that
PR would not fix the generic case, where minimal symbol info is not
available. ]

Fix this by preferring defs over decls when searching in the symbol table.

Build and reg-tested on x86_64-linux.

[ The test-case is a bit simpler than the DWARF example listed above, because
the new variable varval3 that is used is not listed in the minimal symbols, so
there's no need to work around the fallback mechanism to trigger the problem. ]

gdb/ChangeLog:

2019-09-10  Tom de Vries  <tdevries@suse.de>

	PR symtab/24971
	* block.c (best_symbol, better_symbol): New function.
	(block_lookup_symbol_primary): Prefer def over decl.

gdb/testsuite/ChangeLog:

2019-09-10  Tom de Vries  <tdevries@suse.de>

	* gdb.dwarf2/varval.exp: Add decl before def test.

Change-Id: Id92326cb8ef9903b121ef9e320658eb565d0f5a9
---
M gdb/block.c
M gdb/testsuite/gdb.dwarf2/varval.exp
2 files changed, 81 insertions(+), 3 deletions(-)



diff --git a/gdb/block.c b/gdb/block.c
index 5ba44d4..9608194 100644
--- a/gdb/block.c
+++ b/gdb/block.c
@@ -725,6 +725,43 @@
     }
 }
 
+/* Return true if symbol A is the best match possible for DOMAIN.  */
+
+static bool
+best_symbol (struct symbol *a, const domain_enum domain)
+{
+  return (SYMBOL_DOMAIN (a) == domain
+	  && SYMBOL_CLASS (a) != LOC_UNRESOLVED);
+}
+
+/* Return symbol B if it is a better match than symbol A for DOMAIN.
+   Otherwise return A.  */
+
+static struct symbol *
+better_symbol (struct symbol *a, struct symbol *b, const domain_enum domain)
+{
+  if (a == NULL)
+    return b;
+  if (b == NULL)
+    return a;
+
+  if (SYMBOL_DOMAIN (a) == domain
+      && SYMBOL_DOMAIN (b) != domain)
+    return a;
+  if (SYMBOL_DOMAIN (b) == domain
+      && SYMBOL_DOMAIN (a) != domain)
+    return b;
+
+  if (SYMBOL_CLASS (a) != LOC_UNRESOLVED
+      && SYMBOL_CLASS (b) == LOC_UNRESOLVED)
+    return a;
+  if (SYMBOL_CLASS (b) != LOC_UNRESOLVED
+      && SYMBOL_CLASS (a) == LOC_UNRESOLVED)
+    return b;
+
+  return a;
+}
+
 /* See block.h.  */
 
 struct symbol *
@@ -746,7 +783,34 @@
        sym != NULL;
        sym = mdict_iter_match_next (lookup_name, &mdict_iter))
     {
-      if (SYMBOL_DOMAIN (sym) == domain)
+      /* With the fix for PR gcc/debug/91507, we get for:
+	 ...
+	 extern char *zzz[];
+	 char *zzz[ ] = {
+	   "abc",
+	   "cde"
+	 };
+	 ...
+	 DWARF which will result in two entries in the symbol table, a decl
+	 with type char *[] and a def with type char *[2].
+
+	 If we return the decl here, we don't get the value of zzz:
+	 ...
+	 $ gdb a.spec.out -batch -ex "p zzz"
+	 $1 = 0x601030 <zzz>
+	 ...
+	 because we're returning the symbol without location information, and
+	 because the fallback that uses the address from the minimal symbols
+	 doesn't work either because the type of the decl does not specify a
+	 size.
+
+	 To fix this, we prefer def over decl in best_symbol and
+	 better_symbol.
+
+	 In absence of the gcc fix, both def and decl have type char *[], so
+	 the only option to make this work is improve the fallback to use the
+	 size of the minimal symbol.  Filed as PR exp/24989.  */
+      if (best_symbol (sym, domain))
 	return sym;
 
       /* This is a bit of a hack, but symbol_matches_domain might ignore
@@ -755,7 +819,7 @@
 	 exactly the same domain.  PR 16253.  */
       if (symbol_matches_domain (SYMBOL_LANGUAGE (sym),
 				 SYMBOL_DOMAIN (sym), domain))
-	other = sym;
+	other = better_symbol (other, sym, domain);
     }
 
   return other;
diff --git a/gdb/testsuite/gdb.dwarf2/varval.exp b/gdb/testsuite/gdb.dwarf2/varval.exp
index 5945910..0a1b179 100644
--- a/gdb/testsuite/gdb.dwarf2/varval.exp
+++ b/gdb/testsuite/gdb.dwarf2/varval.exp
@@ -55,7 +55,7 @@
 		    var_b_label var_c_label var_p_label var_bad_label \
 		    varval_label var_s_label var_untyped_label \
 		    var_a_abstract_label var_a_concrete_label \
-		    varval2_label
+		    varval2_label varval3_def_label varval3_decl_label
 
 		set int_size [get_sizeof "int" -1]
 
@@ -171,6 +171,19 @@
 		    {DW_AT_location {DW_OP_addr [gdb_target_symbol "var_b"]} SPECIAL_expr}
 		}
 
+		varval3_decl_label: DW_TAG_variable {
+		    {DW_AT_name "varval3"}
+		    {DW_AT_type :${int_label}}
+		    {DW_AT_external 1 DW_FORM_flag}
+		    {DW_AT_declaration 1 DW_FORM_flag}
+		}
+		varval3_def_label: DW_TAG_variable {
+		    {DW_AT_name "varval3"}
+		    {DW_AT_external 1 DW_FORM_flag}
+		    {DW_AT_type :${int_label}}
+		    {DW_AT_location {DW_OP_addr [gdb_target_symbol "var_a"]} SPECIAL_expr}
+		}
+
 		DW_TAG_subprogram {
 		    {MACRO_AT_func { "main" "${srcdir}/${subdir}/${srcfile}" }}
 		    {DW_AT_type :${int_label}}
@@ -290,6 +303,7 @@
 
 gdb_test "print varval" "= 8"
 gdb_test "print varval2" "= 8"
+gdb_test "print varval3" "= 8"
 gdb_test "print constval" "= 53"
 gdb_test "print mixedval" "= 42"
 gdb_test "print pointerval" "= \\(int \\*\\) $hex <var_b>"


  parent reply	other threads:[~2019-10-23 16:27 UTC|newest]

Thread overview: 20+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
     [not found] <gerrit.1571043046000.Id92326cb8ef9903b121ef9e320658eb565d0f5a9@gnutoolchain-gerrit.osci.io>
2019-10-14 16:35 ` Change in binutils-gdb[master]: " Tom de Vries (Code Review)
2019-10-18 13:31 ` [review] " Luis Machado (Code Review)
2019-10-18 16:44 ` Tom de Vries (Code Review)
2019-10-23 16:27 ` Tom de Vries (Code Review) [this message]
2019-10-23 16:29 ` [review v3] " Tom de Vries (Code Review)
2019-10-31 15:50 ` Tom Tromey (Code Review)
2019-11-04  1:11 ` Joel Brobecker (Code Review)
2019-11-07 14:27 ` Tom de Vries (Code Review)
2019-11-07 15:30 ` Tom de Vries (Code Review)
2019-11-11 16:45 ` Joel Brobecker (Code Review)
2019-11-14 16:38 ` Tom de Vries (Code Review)
2019-11-21 13:34 ` Tom Tromey (Code Review)
2019-11-22 10:24 ` Tom de Vries (Code Review)
2019-11-22 11:02 ` Tom de Vries (Code Review)
2019-11-22 12:17 ` [review v4] " Tom de Vries (Code Review)
2019-11-22 12:19 ` Tom de Vries (Code Review)
2019-12-05 20:29 ` Tom Tromey (Code Review)
2019-12-06 17:53 ` [pushed] " Sourceware to Gerrit sync (Code Review)
2019-12-06 17:53 ` Sourceware to Gerrit sync (Code Review)
2019-12-09  6:29 ` [review v5] " Tom de Vries (Code Review)

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20191023162717.844292192E@gnutoolchain-gerrit.osci.io \
    --to=gerrit@gnutoolchain-gerrit.osci.io \
    --cc=andrew.burgess@embecosm.com \
    --cc=gdb-patches@sourceware.org \
    --cc=luis.machado@linaro.org \
    --cc=tdevries@suse.de \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox