[PATCH2,PR,gdb/18071] aarch64: "info addr" command can't resolve TLS variables

Message ID 1522101252-114393-1-git-send-email-weimin.pan@oracle.com
State New
Headers show
Series
  • [PATCH2,PR,gdb/18071] aarch64: "info addr" command can't resolve TLS variables
Related show

Commit Message

Weimin Pan March 26, 2018, 9:54 p.m.
Running the test case with upstream gdb shows two failures:

(1) Receiving different error messages when printing TLS variable before
    program runs - because the ARM compiler does not emit dwarf attribute
    DW_AT_location for TLS, the result is expected and the baseline may
    need to be changed for aarch64.

(2) Using "info address" command on C++ static TLS object resulted in
    "symbol unresolved" error - below is a snippet from the test case:

class K {
 public:
  static __thread int another_thread_local;
};

__thread int K::another_thread_local;

(gdb) info address K::another_thread_local
Symbol "K::another_thread_local" is unresolved.

This patch contains fix for (2).

Function info_address_command() handles the "info address" command and
calls lookup_minimal_symbol_and_objfile() to find sym's symbol entry in
minimal symbol table if SYMBOL_COMPUTED_OPS (sym) is false. Problem is
that function lookup_minimal_symbol_and_objfile() only looked up an
objfile's minsym ordinary hash table, not its demangled hash table, which
was the reason why the C++ name was not found.

The fix is to replace the lookup_minimal_symbol_and_objfile() call with
the lookup_bound_minimal_symbol() call, which looks up entries in both
minsym's hash tables, via lookup_minimal_symbol(), to find symbol entry 
that's associated with the demangled name.

Tested in both aarch64-linux-gnu and amd64-linux-gnu. No regressions.
---
 gdb/ChangeLog  |    8 ++++++++
 gdb/printcmd.c |    2 +-
 2 files changed, 9 insertions(+), 1 deletions(-)

-- 
1.7.1

Comments

Simon Marchi March 27, 2018, 3:31 a.m. | #1
On 2018-03-26 17:54, Weimin Pan wrote:
> Running the test case with upstream gdb shows two failures:

> 

> (1) Receiving different error messages when printing TLS variable 

> before

>     program runs - because the ARM compiler does not emit dwarf 

> attribute

>     DW_AT_location for TLS, the result is expected and the baseline may

>     need to be changed for aarch64.

> 

> (2) Using "info address" command on C++ static TLS object resulted in

>     "symbol unresolved" error - below is a snippet from the test case:

> 

> class K {

>  public:

>   static __thread int another_thread_local;

> };

> 

> __thread int K::another_thread_local;

> 

> (gdb) info address K::another_thread_local

> Symbol "K::another_thread_local" is unresolved.

> 

> This patch contains fix for (2).

> 

> Function info_address_command() handles the "info address" command and

> calls lookup_minimal_symbol_and_objfile() to find sym's symbol entry in

> minimal symbol table if SYMBOL_COMPUTED_OPS (sym) is false. Problem is

> that function lookup_minimal_symbol_and_objfile() only looked up an

> objfile's minsym ordinary hash table, not its demangled hash table, 

> which

> was the reason why the C++ name was not found.

> 

> The fix is to replace the lookup_minimal_symbol_and_objfile() call with

> the lookup_bound_minimal_symbol() call, which looks up entries in both

> minsym's hash tables, via lookup_minimal_symbol(), to find symbol entry

> that's associated with the demangled name.


Hi Weimin,

I don't know if it was clear, but I pushed this patch already:

https://sourceware.org/git/gitweb.cgi?p=binutils-gdb.git;a=commit;h=bce02d8884d6baa72c537d0d7c59f924cb290799

Removing lookup_minimal_symbol_and_objfile and replacing it with 
lookup_bound_minimal_symbol should be done as a separate patch (with a 
relevant title that describes the change).  Can you do that?  There are 
just a few callers of lookup_minimal_symbol_and_objfile that will need 
to be updated.

Simon
Weimin Pan March 27, 2018, 7:55 p.m. | #2
On 3/26/2018 8:31 PM, Simon Marchi wrote:
> On 2018-03-26 17:54, Weimin Pan wrote:

>> Running the test case with upstream gdb shows two failures:

>>

>> (1) Receiving different error messages when printing TLS variable before

>>     program runs - because the ARM compiler does not emit dwarf 

>> attribute

>>     DW_AT_location for TLS, the result is expected and the baseline may

>>     need to be changed for aarch64.

>>

>> (2) Using "info address" command on C++ static TLS object resulted in

>>     "symbol unresolved" error - below is a snippet from the test case:

>>

>> class K {

>>  public:

>>   static __thread int another_thread_local;

>> };

>>

>> __thread int K::another_thread_local;

>>

>> (gdb) info address K::another_thread_local

>> Symbol "K::another_thread_local" is unresolved.

>>

>> This patch contains fix for (2).

>>

>> Function info_address_command() handles the "info address" command and

>> calls lookup_minimal_symbol_and_objfile() to find sym's symbol entry in

>> minimal symbol table if SYMBOL_COMPUTED_OPS (sym) is false. Problem is

>> that function lookup_minimal_symbol_and_objfile() only looked up an

>> objfile's minsym ordinary hash table, not its demangled hash table, 

>> which

>> was the reason why the C++ name was not found.

>>

>> The fix is to replace the lookup_minimal_symbol_and_objfile() call with

>> the lookup_bound_minimal_symbol() call, which looks up entries in both

>> minsym's hash tables, via lookup_minimal_symbol(), to find symbol entry

>> that's associated with the demangled name.

>

> Hi Weimin,

>

> I don't know if it was clear, but I pushed this patch already:

>

> https://sourceware.org/git/gitweb.cgi?p=binutils-gdb.git;a=commit;h=bce02d8884d6baa72c537d0d7c59f924cb290799 

>

>

> Removing lookup_minimal_symbol_and_objfile and replacing it with 

> lookup_bound_minimal_symbol should be done as a separate patch (with a 

> relevant title that describes the change).  Can you do that?  There 

> are just a few callers of lookup_minimal_symbol_and_objfile that will 

> need to be updated.

>


Hi Simon,

Yes, I will do another patch  get rid of all 
lookup_minimal_symbol_and_objfile calls.

Weimin

> Simon

Patch

diff --git a/gdb/ChangeLog b/gdb/ChangeLog
index fe4ae9f..5faa37b 100644
--- a/gdb/ChangeLog
+++ b/gdb/ChangeLog
@@ -1,3 +1,11 @@ 
+2018-03-22  Weimin Pan  <weimin.pan@oracle.com>
+
+	PR gdb/18071:
+	* printcmd.c (info_address_command): Replace
+	lookup_minimal_symbol_and_objfile call with
+	lookup_bound_minimal_symbol call to find symbol entry
+	of a demangled name.
+
 2018-03-19  Pedro Alves  <palves@redhat.com>
 	    Tom Tromey  <tom@tromey.com>
 
diff --git a/gdb/printcmd.c b/gdb/printcmd.c
index dd81d8f..06038cd 100644
--- a/gdb/printcmd.c
+++ b/gdb/printcmd.c
@@ -1564,7 +1564,7 @@  info_address_command (const char *exp, int from_tty)
       {
 	struct bound_minimal_symbol msym;
 
-	msym = lookup_minimal_symbol_and_objfile (SYMBOL_LINKAGE_NAME (sym));
+	msym = lookup_bound_minimal_symbol (SYMBOL_LINKAGE_NAME (sym));
 	if (msym.minsym == NULL)
 	  printf_filtered ("unresolved");
 	else