Re: [RFA] port simple gdb.threads/schedlock.c test fix to branch

Re: [RFA] port simple gdb.threads/schedlock.c test fix to branch

[Michael Elizabeth Chastain]

> This is OK for the branch.  I think I meant to do it at the time and
> dropped the ball. I should also check that the lin-lwp fix for
> schedlock.exp made the branch...

From the black box point of view, gdb HEAD and gdb 5.3 branch are
behaving very differently with schedlock.exp.  This is with all the
same compilers and binutils and yada.  gdb HEAD gives me 2 FAILs
consistently, and gdb 5.3 branch gives me 8 FAILs most of the time
and 4 FAILs some of the time.

Gory details below.

My version is from 2002-11-25, a week ago, and you checked in one
fix to lin-lwp.c since then, to call linux_proc_xfer_memory.
Is that the fix that lin-lwp.c needs?

Michael C

  schedlock.exp FAILs

    gdb.thread/schedlock.exp: thread 0 ran (didn't run)
    gdb.thread/schedlock.exp: thread 1 ran (didn't run)
      ++ happened in all gdb HEAD configurations (130 of 130)
      ++ did not happen in gdb gdb_5_3-branch
      ++ script did not exist in gdb 5.2.1

    gdb.thread/schedlock.exp: thread 0 ran (didn't run)
    gdb.thread/schedlock.exp: thread 1 ran (didn't run)
    gdb.threads/schedlock.exp: continue with lock does not change thread (switched to thread 3)
    gdb.threads/schedlock.exp: other thread 3 didn't run (ran)
    gbd.threads/schedlock.exp: other thread 4 didn't run (ran)
    gdb.threads/schedlock.exp: step with lock does not change thread (switched to thread 3)
    gdb.threads/schedlock.exp: current thread stepped locked (didn't run)
    gdb.threads/schedlock.exp: other thread 3 didn't run (stepping) (ran)
      ++ never happened in gdb HEAD
      ++ happened in most gdb gdb_5_3-branch configurations (99 of 130)
      ++ script did not exist in gdb 5.2.1
      ++ not correlated with gcc version or binutils version

    gdb.thread/schedlock.exp: thread 0 ran (didn't run)
    gdb.thread/schedlock.exp: thread 1 ran (didn't run)
    gdb.threads/schedlock.exp: other thread 3 didn't run (ran)
    gbd.threads/schedlock.exp: other thread 4 didn't run (ran)
      ++ never happened in gdb HEAD
      ++ happened in some gdb gdb_5_3-branch configurations (31 of 130)
      ++ script did not exist in gdb 5.2.1
      ++ not correlated with gcc version or binutils version

Re: [RFA] port simple gdb.threads/schedlock.c test fix to branch

[Daniel Jacobowitz]

On Mon, Dec 02, 2002 at 11:19:24PM -0600, Michael Elizabeth Chastain wrote:
> > This is OK for the branch.  I think I meant to do it at the time and
> > dropped the ball. I should also check that the lin-lwp fix for
> > schedlock.exp made the branch...
> 
> >From the black box point of view, gdb HEAD and gdb 5.3 branch are
> behaving very differently with schedlock.exp.  This is with all the
> same compilers and binutils and yada.  gdb HEAD gives me 2 FAILs
> consistently, and gdb 5.3 branch gives me 8 FAILs most of the time
> and 4 FAILs some of the time.
> 
> Gory details below.
> 
> My version is from 2002-11-25, a week ago, and you checked in one
> fix to lin-lwp.c since then, to call linux_proc_xfer_memory.
> Is that the fix that lin-lwp.c needs?

No.  It should be, um...
2002-10-31  Daniel Jacobowitz  <drow@mvista.com>

        * lin-lwp.c (lin_lwp_resume): Remove resume_all test for !step.


-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer

[RFA] port simple gdb.threads/schedlock.c test fix to branch

[Michael Elizabeth Chastain]

I would like to port this fix from mainline to 5.3 branch:

  http://sources.redhat.com/ml/gdb-patches/2002-10/msg00454.html

  2002-10-22  Daniel Jacobowitz  <drow@mvista.com>

        * gdb.threads/schedlock.c (args): Make unsigned.

Note that this is a test program fix, not a gdb fix.

This fix improves the results from schedlock.exp, making them more
accurate and reproducible.  Right now, I'm getting random failures
in schedlock.exp on the branch, and it takes work to determine that
the failures really are random and not correlated with different
versions of gcc and binutils.

I tested this patch in my test bed with 66 different configurations
(on the 5.3 branch) and it works fine.  And, of course, it's been
working in mainline just fine.

I think this is obvious, but since I've been away for a while,
I just want to check first.

Okay to apply to the 5.3 branch?

Michael C

===

2002-12-02  Michael Chastain  <mec@shout.net>

	From mainline:
	2002-10-22  Daniel Jacobowitz  <drow@mvista.com>
        * gdb.threads/schedlock.c (args): Make unsigned.

Index: testsuite/gdb.threads/schedlock.c
===================================================================
RCS file: /cvs/src/src/gdb/testsuite/gdb.threads/schedlock.c,v
retrieving revision 1.1
diff -u -p -r1.1 schedlock.c
--- testsuite/gdb.threads/schedlock.c	23 Aug 2002 20:32:02 -0000	1.1
+++ testsuite/gdb.threads/schedlock.c	23 Oct 2002 03:21:13 -0000
@@ -7,7 +7,7 @@ void *thread_function(void *arg); /* Poi
 
 #define NUM 5
 
-int args[NUM+1];
+unsigned int args[NUM+1];
 
 int main() {
     int res;

Re: [RFA] port simple gdb.threads/schedlock.c test fix to branch

[Daniel Jacobowitz]

On Mon, Dec 02, 2002 at 10:20:13PM -0600, Michael Elizabeth Chastain wrote:
> I would like to port this fix from mainline to 5.3 branch:
> 
>   http://sources.redhat.com/ml/gdb-patches/2002-10/msg00454.html
> 
>   2002-10-22  Daniel Jacobowitz  <drow@mvista.com>
> 
>         * gdb.threads/schedlock.c (args): Make unsigned.
> 
> Note that this is a test program fix, not a gdb fix.
> 
> This fix improves the results from schedlock.exp, making them more
> accurate and reproducible.  Right now, I'm getting random failures
> in schedlock.exp on the branch, and it takes work to determine that
> the failures really are random and not correlated with different
> versions of gcc and binutils.
> 
> I tested this patch in my test bed with 66 different configurations
> (on the 5.3 branch) and it works fine.  And, of course, it's been
> working in mainline just fine.
> 
> I think this is obvious, but since I've been away for a while,
> I just want to check first.
> 
> Okay to apply to the 5.3 branch?

<obvious hat>

This is OK for the branch.  I think I meant to do it at the time and
dropped the ball. I should also check that the lin-lwp fix for
schedlock.exp made the branch...

Caveat: Andrew had closed the branch for 5.2.91; I don't know if he's
opened it up again yet or how strictly we're frozen.  Andrew?

> 
> Michael C
> 
> ===
> 
> 2002-12-02  Michael Chastain  <mec@shout.net>
> 
> 	From mainline:
> 	2002-10-22  Daniel Jacobowitz  <drow@mvista.com>
>         * gdb.threads/schedlock.c (args): Make unsigned.
> 
> Index: testsuite/gdb.threads/schedlock.c
> ===================================================================
> RCS file: /cvs/src/src/gdb/testsuite/gdb.threads/schedlock.c,v
> retrieving revision 1.1
> diff -u -p -r1.1 schedlock.c
> --- testsuite/gdb.threads/schedlock.c	23 Aug 2002 20:32:02 -0000	1.1
> +++ testsuite/gdb.threads/schedlock.c	23 Oct 2002 03:21:13 -0000
> @@ -7,7 +7,7 @@ void *thread_function(void *arg); /* Poi
>  
>  #define NUM 5
>  
> -int args[NUM+1];
> +unsigned int args[NUM+1];
>  
>  int main() {
>      int res;
> 
> 
> 
> 

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer