Fix gdb.python/py-thrhandle.exp failures for -m32 multilib

Message ID 20190726015013.3970-1-kevinb@redhat.com
State New
Headers show
Series
  • Fix gdb.python/py-thrhandle.exp failures for -m32 multilib
Related show

Commit Message

Kevin Buettner July 26, 2019, 1:50 a.m.
This patch fixes the following failures when testing with
"target_board unix/-m32" using a x86_64-pc-linux-gnu native GDB.

FAIL: gdb.python/py-thrhandle.exp: print thread for bogus handle thrs[3]
FAIL: gdb.python/py-thrhandle.exp: print thread for bogus handle thrs[4]
FAIL: gdb.python/py-thrhandle.exp: print thread id for thrs[0]
FAIL: gdb.python/py-thrhandle.exp: print thread id for thrs[1]
FAIL: gdb.python/py-thrhandle.exp: print thread id for thrs[2]
FAIL: gdb.python/py-thrhandle.exp: thread 0: fetch thread handle from thread
FAIL: gdb.python/py-thrhandle.exp: thread 0: verify that handles are the same
FAIL: gdb.python/py-thrhandle.exp: thread 1: fetch thread handle from thread
FAIL: gdb.python/py-thrhandle.exp: thread 1: verify that handles are the same
FAIL: gdb.python/py-thrhandle.exp: thread 2: fetch thread handle from thread
FAIL: gdb.python/py-thrhandle.exp: thread 2: verify that handles are the same

I've written it so that it might work for other 64-bit host / 32-bit target
combos too.

gdb/ChangeLog:

	* linux-thread-db.c (thread_db_target::thread_handle_to_thread_info):
	Add case for debugging 32-bit target on 64-bit host.  Revise
	comment.
---
 gdb/linux-thread-db.c | 19 ++++++++++++-------
 1 file changed, 12 insertions(+), 7 deletions(-)

-- 
2.21.0

Comments

Tom Tromey July 26, 2019, 9:01 p.m. | #1
>>>>> "Kevin" == Kevin Buettner <kevinb@redhat.com> writes:


Kevin> I've written it so that it might work for other 64-bit host / 32-bit target
Kevin> combos too.

Kevin> gdb/ChangeLog:

Kevin> 	* linux-thread-db.c (thread_db_target::thread_handle_to_thread_info):
Kevin> 	Add case for debugging 32-bit target on 64-bit host.  Revise
Kevin> 	comment.

Looks reasonable to me.

Tom
Kevin Buettner July 27, 2019, 8:54 p.m. | #2
On Fri, 26 Jul 2019 15:01:33 -0600
Tom Tromey <tom@tromey.com> wrote:

> >>>>> "Kevin" == Kevin Buettner <kevinb@redhat.com> writes:  

> 

> Kevin> I've written it so that it might work for other 64-bit host / 32-bit target

> Kevin> combos too.  

> 

> Kevin> gdb/ChangeLog:  

> 

> Kevin> 	* linux-thread-db.c (thread_db_target::thread_handle_to_thread_info):

> Kevin> 	Add case for debugging 32-bit target on 64-bit host.  Revise

> Kevin> 	comment.  

> 

> Looks reasonable to me.


Thanks for looking it over.

Pushed.

Kevin

Patch

diff --git a/gdb/linux-thread-db.c b/gdb/linux-thread-db.c
index 950be9ca5a..1d32e9195c 100644
--- a/gdb/linux-thread-db.c
+++ b/gdb/linux-thread-db.c
@@ -1672,16 +1672,21 @@  thread_db_target::thread_handle_to_thread_info (const gdb_byte *thread_handle,
 {
   thread_t handle_tid;
 
-  /* Thread handle sizes must match in order to proceed.  We don't use an
-     assert here because the resulting internal error will cause GDB to
-     exit.  This isn't necessarily an internal error due to the possibility
-     of garbage being passed as the thread handle via the python interface.  */
-  if (handle_len != sizeof (handle_tid))
+  /* When debugging a 32-bit target from a 64-bit host, handle_len
+     will be 4 and sizeof (handle_tid) will be 8.  This requires
+     a different cast than the more straightforward case where
+     the sizes are the same.
+
+     Use "--target_board unix/-m32" from a native x86_64 linux build
+     to test the 32/64-bit case.  */
+  if (handle_len == 4 && sizeof (handle_tid) == 8)
+    handle_tid = (thread_t) * (const uint32_t *) thread_handle;
+  else if (handle_len == sizeof (handle_tid))
+    handle_tid = * (const thread_t *) thread_handle;
+  else
     error (_("Thread handle size mismatch: %d vs %zu (from libthread_db)"),
 	   handle_len, sizeof (handle_tid));
 
-  handle_tid = * (const thread_t *) thread_handle;
-
   for (thread_info *tp : inf->non_exited_threads ())
     {
       thread_db_thread_info *priv = get_thread_db_thread_info (tp);