[rfc target-side break conditions 0/5 v2] General info

[rfc target-side break conditions 0/5 v2] General info

[Luis Gustavo]

Hi,

This is an updated series of patches to implement target-side breakpoint 
condition evaluation.

Most of the changes are in part 1, 3 and 5. The other parts had only 
minor changes.

The current testsuite seems to be a good exercise for this feature when 
ran in gdbserver mode. I can provide additional tests if required.

Part 1: Documentation bits
Part 2: Protocol and feature changes
Part 3: GDB-specific changes
Part 4: GDBServer tracepoint/agent expression common code cleanup
Part 5: GDBServer breakpoint condition evaluation enablement

Luis

[ping] [rfc target-side break conditions 0/5 v2] General info

[Luis Gustavo]

Ping this updated series.

On 01/27/2012 06:32 PM, Luis Gustavo wrote:
> Hi,
>
> This is an updated series of patches to implement target-side breakpoint
> condition evaluation.
>
> Most of the changes are in part 1, 3 and 5. The other parts had only
> minor changes.
>
> The current testsuite seems to be a good exercise for this feature when
> ran in gdbserver mode. I can provide additional tests if required.
>
> Part 1: Documentation bits
> Part 2: Protocol and feature changes
> Part 3: GDB-specific changes
> Part 4: GDBServer tracepoint/agent expression common code cleanup
> Part 5: GDBServer breakpoint condition evaluation enablement
>
> Luis
>

Re: [rfc target-side break conditions 0/5 v2] General info

[Stan Shebs]

On 1/27/12 12:32 PM, Luis Gustavo wrote:
> Hi,
>
> This is an updated series of patches to implement target-side 
> breakpoint condition evaluation.

Overall this is looking pretty good, I just have a few nitpicks of my 
own - fix those and it all should be fine to go in.

>
> The current testsuite seems to be a good exercise for this feature 
> when ran in gdbserver mode. I can provide additional tests if required.

I would like to see at least one test of the preference setting that 
runs in native mode and comes back with "not supported".  I think this 
is something that people will imagine is possible for native debugging, 
and maybe someday it will be, but in the meantime we want to be sure 
that the native config clearly reports that the setting is not going to 
have any effect.

Stan

Re: [rfc target-side break conditions 0/5 v2] General info

[Luis Gustavo]

On 02/07/2012 07:32 PM, Stan Shebs wrote:
> On 1/27/12 12:32 PM, Luis Gustavo wrote:
>> Hi,
>>
>> This is an updated series of patches to implement target-side
>> breakpoint condition evaluation.
>
> Overall this is looking pretty good, I just have a few nitpicks of my
> own - fix those and it all should be fine to go in.
>
>>
>> The current testsuite seems to be a good exercise for this feature
>> when ran in gdbserver mode. I can provide additional tests if required.
>
> I would like to see at least one test of the preference setting that
> runs in native mode and comes back with "not supported". I think this is
> something that people will imagine is possible for native debugging, and
> maybe someday it will be, but in the meantime we want to be sure that
> the native config clearly reports that the setting is not going to have
> any effect.

I'll attach the testcase to a new mail...

Luis

Re: [rfc target-side break conditions 0/5 v2] General info

[Luis Gustavo]

On 01/27/2012 06:32 PM, Luis Gustavo wrote:
> Hi,
>
> This is an updated series of patches to implement target-side breakpoint
> condition evaluation.
>
> Most of the changes are in part 1, 3 and 5. The other parts had only
> minor changes.
>
> The current testsuite seems to be a good exercise for this feature when
> ran in gdbserver mode. I can provide additional tests if required.
>
> Part 1: Documentation bits
> Part 2: Protocol and feature changes
> Part 3: GDB-specific changes
> Part 4: GDBServer tracepoint/agent expression common code cleanup
> Part 5: GDBServer breakpoint condition evaluation enablement
>
> Luis
>

I've checked this whole series in now.

Thanks!
Luis

General regressions in gdbserver mode [Re: [rfc target-side break conditions 0/5 v2] General info]

[Jan Kratochvil]

Hi Luis,

On Fri, 24 Feb 2012 16:25:03 +0100, Luis Gustavo wrote:
> I've checked this whole series in now.

while thanks for this new feature the testsuite results broke in
	http://sourceware.org/gdb/wiki/TestingGDB#Testing_gdbserver_in_a_native_configuration
mode such as:

(gdb) continue
Continuing.
Breakpoint 1, main (argc=1, argv=0x7fffffffdbd8, envp=0x7fffffffdbe8) at ./gdb.base/break.c:89
89          if (argc == 12345) {  /* an unlikely value < 2^16, in case uninited */ /* set breakpoint 6 here */
(gdb) PASS: gdb.base/condbreak.exp: rerun to main

->

(gdb) continue
Continuing.
Found breakpoint condition.
Found breakpoint condition.
Found breakpoint condition.
Remote connection closed
(gdb) FAIL: gdb.base/condbreak.exp: rerun to main

I do not have exact results today (offtopic) but the regression spans across
the testsuite in general.


Thanks,
Jan

Re: General regressions in gdbserver mode [Re: [rfc target-side break conditions 0/5 v2] General info]

[Luis Gustavo]

Hi Jan,

On 02/25/2012 12:20 PM, Jan Kratochvil wrote:
> Hi Luis,
>
> On Fri, 24 Feb 2012 16:25:03 +0100, Luis Gustavo wrote:
>> I've checked this whole series in now.
>
> while thanks for this new feature the testsuite results broke in
> 	http://sourceware.org/gdb/wiki/TestingGDB#Testing_gdbserver_in_a_native_configuration
> mode such as:
>
> (gdb) continue
> Continuing.
> Breakpoint 1, main (argc=1, argv=0x7fffffffdbd8, envp=0x7fffffffdbe8) at ./gdb.base/break.c:89
> 89          if (argc == 12345) {  /* an unlikely value<  2^16, in case uninited */ /* set breakpoint 6 here */
> (gdb) PASS: gdb.base/condbreak.exp: rerun to main
>
> ->
>
> (gdb) continue
> Continuing.
> Found breakpoint condition.
> Found breakpoint condition.
> Found breakpoint condition.
> Remote connection closed
> (gdb) FAIL: gdb.base/condbreak.exp: rerun to main
>
> I do not have exact results today (offtopic) but the regression spans across
> the testsuite in general.
>
>
> Thanks,
> Jan

What kind of setup you're using? I have full passes for this testcase 
using the native gdbserver mode on x86-32 for example.

Breakpoint 1, main (argc=1, argv=0xbfffdef4, envp=0xbfffdefc) at 
../../../gdb/gdb/testsuite/gdb.base/break.c:89^M
89          if (argc == 12345) {  /* an unlikely value < 2^16, in case 
uninited */ /* set breakpoint 6 here */^M
(gdb) PASS: gdb.base/condbreak.exp: rerun to main
continue^M
Continuing.^M
^M
Breakpoint 5, main (argc=1, argv=0xbfffdef4, envp=0xbfffdefc) at 
../../../gdb/gdb/testsuite/gdb.base/break.c:93^M
93          printf ("%d\n", factorial (atoi ("6")));  /* set breakpoint 
1 here */^M
(gdb) PASS: gdb.base/condbreak.exp: run until breakpoint set at a line 
number
Found breakpoint condition.^M
Found breakpoint condition.^M
Found breakpoint condition.^M
Found breakpoint condition.^M
Found breakpoint condition.^M
Found breakpoint condition.^M
720^M
continue^M
Continuing.^M

I wonder if it's something wrong with agent expression parsing.

Luis

Re: General regressions in gdbserver mode [Re: [rfc target-side break conditions 0/5 v2] General info]

[Jan Kratochvil]

On Sat, 25 Feb 2012 16:39:55 +0100, Luis Gustavo wrote:
> What kind of setup you're using?

Fedora 17 x86_64.  It does not crash for me without -lmcheck but it does crash
for gdbserver linked with -lmcheck, gdbserver dumps core then.

With valgrind gdbserver prints:

Invalid read of size 8
   at: clear_gdb_breakpoint_conditions (mem-break.c:741)
   by: delete_gdb_breakpoint_at (mem-break.c:715)
   by: x86_remove_point (linux-x86-low.c:593)
   by: linux_remove_point (linux-low.c:4463)
   by: process_serial_event (server.c:3207)
   by: handle_serial_event (server.c:3340)
   by: handle_file_event (event-loop.c:489)
   by: process_event (event-loop.c:244)
   by: start_event_loop (event-loop.c:607)
   by: main (server.c:2773)
 Address 0x56242c8 is 8 bytes inside a block of size 16 free'd
   at: free (vg_replace_malloc.c:427)
   by: clear_gdb_breakpoint_conditions (mem-break.c:740)
   by: delete_gdb_breakpoint_at (mem-break.c:715)
   by: x86_remove_point (linux-x86-low.c:593)
   by: linux_remove_point (linux-low.c:4463)
   by: process_serial_event (server.c:3207)
   by: handle_serial_event (server.c:3340)
   by: handle_file_event (event-loop.c:489)
   by: process_event (event-loop.c:244)
   by: start_event_loop (event-loop.c:607)
   by: main (server.c:2773)


Thanks,
Jan

Re: General regressions in gdbserver mode [Re: [rfc target-side break conditions 0/5 v2] General info]

[Luis Gustavo]

On 02/25/2012 03:06 PM, Jan Kratochvil wrote:
> On Sat, 25 Feb 2012 16:39:55 +0100, Luis Gustavo wrote:
>> What kind of setup you're using?
>
> Fedora 17 x86_64.  It does not crash for me without -lmcheck but it does crash
> for gdbserver linked with -lmcheck, gdbserver dumps core then.
>
> With valgrind gdbserver prints:
>
> Invalid read of size 8
>     at: clear_gdb_breakpoint_conditions (mem-break.c:741)
>     by: delete_gdb_breakpoint_at (mem-break.c:715)
>     by: x86_remove_point (linux-x86-low.c:593)
>     by: linux_remove_point (linux-low.c:4463)
>     by: process_serial_event (server.c:3207)
>     by: handle_serial_event (server.c:3340)
>     by: handle_file_event (event-loop.c:489)
>     by: process_event (event-loop.c:244)
>     by: start_event_loop (event-loop.c:607)
>     by: main (server.c:2773)
>   Address 0x56242c8 is 8 bytes inside a block of size 16 free'd
>     at: free (vg_replace_malloc.c:427)
>     by: clear_gdb_breakpoint_conditions (mem-break.c:740)
>     by: delete_gdb_breakpoint_at (mem-break.c:715)
>     by: x86_remove_point (linux-x86-low.c:593)
>     by: linux_remove_point (linux-low.c:4463)
>     by: process_serial_event (server.c:3207)
>     by: handle_serial_event (server.c:3340)
>     by: handle_file_event (event-loop.c:489)
>     by: process_event (event-loop.c:244)
>     by: start_event_loop (event-loop.c:607)
>     by: main (server.c:2773)
>
>
> Thanks,
> Jan

Thanks for the info. Let me check this.

Luis

Re: General regressions in gdbserver mode [Re: [rfc target-side break conditions 0/5 v2] General info]

[Luis Gustavo]

[-- Attachment #1: Type: text/plain, Size: 1549 bytes --]

On 02/25/2012 03:06 PM, Jan Kratochvil wrote:
> On Sat, 25 Feb 2012 16:39:55 +0100, Luis Gustavo wrote:
>> What kind of setup you're using?
>
> Fedora 17 x86_64.  It does not crash for me without -lmcheck but it does crash
> for gdbserver linked with -lmcheck, gdbserver dumps core then.
>
> With valgrind gdbserver prints:
>
> Invalid read of size 8
>     at: clear_gdb_breakpoint_conditions (mem-break.c:741)
>     by: delete_gdb_breakpoint_at (mem-break.c:715)
>     by: x86_remove_point (linux-x86-low.c:593)
>     by: linux_remove_point (linux-low.c:4463)
>     by: process_serial_event (server.c:3207)
>     by: handle_serial_event (server.c:3340)
>     by: handle_file_event (event-loop.c:489)
>     by: process_event (event-loop.c:244)
>     by: start_event_loop (event-loop.c:607)
>     by: main (server.c:2773)
>   Address 0x56242c8 is 8 bytes inside a block of size 16 free'd
>     at: free (vg_replace_malloc.c:427)
>     by: clear_gdb_breakpoint_conditions (mem-break.c:740)
>     by: delete_gdb_breakpoint_at (mem-break.c:715)
>     by: x86_remove_point (linux-x86-low.c:593)
>     by: linux_remove_point (linux-low.c:4463)
>     by: process_serial_event (server.c:3207)
>     by: handle_serial_event (server.c:3340)
>     by: handle_file_event (event-loop.c:489)
>     by: process_event (event-loop.c:244)
>     by: start_event_loop (event-loop.c:607)
>     by: main (server.c:2773)
>
>
> Thanks,
> Jan

Could please try the following on your side?

I reproduced the crash with -lmcheck and this patch fixed it for me.

Thanks,
Luis


[-- Attachment #2: mem_fix.diff --]
[-- Type: text/x-patch, Size: 868 bytes --]

2012-02-25  Luis Machado  <lgustavo@codesourcery.com>

	* mem-break.c (clear_gdb_breakpoint_conditions): Fix de-allocation
	of conditions.

diff --git a/gdb/gdbserver/mem-break.c b/gdb/gdbserver/mem-break.c
index c9a6035..69265cb 100644
--- a/gdb/gdbserver/mem-break.c
+++ b/gdb/gdbserver/mem-break.c
@@ -726,20 +726,19 @@ void
 clear_gdb_breakpoint_conditions (CORE_ADDR addr)
 {
   struct breakpoint *bp = find_gdb_breakpoint_at (addr);
-  struct point_cond_list *cond, **cond_p;
+  struct point_cond_list *cond, *cond_next;
 
   if (bp == NULL || bp->cond_list == NULL)
     return;
 
   cond = bp->cond_list;
-  cond_p = &bp->cond_list->next;
 
   while (cond != NULL)
     {
+      cond_next = cond->next;
       free (cond->cond);
       free (cond);
-      cond = *cond_p;
-      cond_p = &cond->next;
+      cond = cond_next;
     }
 
   bp->cond_list = NULL;

Re: General regressions in gdbserver mode [Re: [rfc target-side break conditions 0/5 v2] General info]

[Jan Kratochvil]

On Sat, 25 Feb 2012 19:24:58 +0100, Luis Gustavo wrote:
> Could please try the following on your side?

Yes, it works for me.

Please check it in, it fixes the regression.


> --- a/gdb/gdbserver/mem-break.c
> +++ b/gdb/gdbserver/mem-break.c
> @@ -726,20 +726,19 @@ void
>  clear_gdb_breakpoint_conditions (CORE_ADDR addr)
>  {
>    struct breakpoint *bp = find_gdb_breakpoint_at (addr);
> -  struct point_cond_list *cond, **cond_p;
> +  struct point_cond_list *cond, *cond_next;
>  
>    if (bp == NULL || bp->cond_list == NULL)
>      return;
>  
>    cond = bp->cond_list;
> -  cond_p = &bp->cond_list->next;
>  
>    while (cond != NULL)
>      {

cond_next could be decllared inside in this block.

> +      cond_next = cond->next;

Isn't missing here also?
	free (cond->cond->bytes);


>        free (cond->cond);
>        free (cond);
> -      cond = *cond_p;
> -      cond_p = &cond->next;
> +      cond = cond_next;
>      }
>  
>    bp->cond_list = NULL;


Thanks,
Jan

Re: General regressions in gdbserver mode [Re: [rfc target-side break conditions 0/5 v2] General info]

[Luis Gustavo]

[-- Attachment #1: Type: text/plain, Size: 1024 bytes --]

Hi Jan,

On 02/25/2012 05:40 PM, Jan Kratochvil wrote:
> On Sat, 25 Feb 2012 19:24:58 +0100, Luis Gustavo wrote:
>> Could please try the following on your side?
>
> Yes, it works for me.
>
> Please check it in, it fixes the regression.
>
>
>> --- a/gdb/gdbserver/mem-break.c
>> +++ b/gdb/gdbserver/mem-break.c
>> @@ -726,20 +726,19 @@ void
>>   clear_gdb_breakpoint_conditions (CORE_ADDR addr)
>>   {
>>     struct breakpoint *bp = find_gdb_breakpoint_at (addr);
>> -  struct point_cond_list *cond, **cond_p;
>> +  struct point_cond_list *cond, *cond_next;
>>
>>     if (bp == NULL || bp->cond_list == NULL)
>>       return;
>>
>>     cond = bp->cond_list;
>> -  cond_p =&bp->cond_list->next;
>>
>>     while (cond != NULL)
>>       {
>
> cond_next could be decllared inside in this block.
>

Fixed.

>> +      cond_next = cond->next;
>
> Isn't missing here also?
> 	free (cond->cond->bytes);
>
>

It is, thanks. GDBserver does not know free_agent_expr currently, maybe 
in the future.

I've checked the following in.

Luis

[-- Attachment #2: mem_fix.diff --]
[-- Type: text/x-patch, Size: 933 bytes --]

2012-02-25  Luis Machado  <lgustavo@codesourcery.com>

	* mem-break.c (clear_gdb_breakpoint_conditions): Fix de-allocation
	of conditions.

diff --git a/gdb/gdbserver/mem-break.c b/gdb/gdbserver/mem-break.c
index c9a6035..6b6b25c 100644
--- a/gdb/gdbserver/mem-break.c
+++ b/gdb/gdbserver/mem-break.c
@@ -726,20 +726,22 @@ void
 clear_gdb_breakpoint_conditions (CORE_ADDR addr)
 {
   struct breakpoint *bp = find_gdb_breakpoint_at (addr);
-  struct point_cond_list *cond, **cond_p;
+  struct point_cond_list *cond;
 
   if (bp == NULL || bp->cond_list == NULL)
     return;
 
   cond = bp->cond_list;
-  cond_p = &bp->cond_list->next;
 
   while (cond != NULL)
     {
+      struct point_cond_list *cond_next;
+
+      cond_next = cond->next;
+      free (cond->cond->bytes);
       free (cond->cond);
       free (cond);
-      cond = *cond_p;
-      cond_p = &cond->next;
+      cond = cond_next;
     }
 
   bp->cond_list = NULL;

Regressions in gdbserver mode #2 [Re: General regressions in gdbserver mode]

[Jan Kratochvil]

On Sat, 25 Feb 2012 20:58:53 +0100, Luis Gustavo wrote:
> I've checked the following in.

Still getting now these regressions in gdbserver non-extended mode:

-PASS: gdb.base/save-bp.exp: info break
+FAIL: gdb.base/save-bp.exp: info break
+FAIL: gdb.cp/mb-templates.exp: initial condition: run to breakpoint (timeout)
+FAIL: gdb.cp/mb-templates.exp: initial condition: run to breakpoint 2 (GDB internal error)
+FAIL: gdb.cp/mb-templates.exp: ...

info break
Num     Type           Disp Enb Address            What
2       breakpoint     keep y   0x0000000000400540 in break_me at ./gdb.base/save-bp.c:21
3       breakpoint     keep y   0x000000000040054a in main at ./gdb.base/save-bp.c:27
4       breakpoint     keep y   0x000000000040054f in main at ./gdb.base/save-bp.c:28 thread 1
        stop only in thread 1
5       breakpoint     keep y   0x000000000040055d in main at ./gdb.base/save-bp.c:31
        stop only if i == 1 (host evals)
6       breakpoint     keep y   0x000000000040056c in main at ./gdb.base/save-bp.c:33
        silent
(gdb) FAIL: gdb.base/save-bp.exp: info break

(I did not check now what is expeted.)

(gdb) continue
Continuing.
ax-gdb.c:560: internal-error: gen_fetch: bad type code
A problem internal to GDB has been detected,
further debugging may prove unreliable.
Quit this debugging session? (y or n) Found breakpoint condition.
FAIL: gdb.cp/mb-templates.exp: initial condition: run to breakpoint (timeout)


Regards,
Jan

Re: Regressions in gdbserver mode #2 [Re: General regressions in gdbserver mode]

[Luis Gustavo]

Hi Jan,


On 02/26/2012 06:54 PM, Jan Kratochvil wrote:
> On Sat, 25 Feb 2012 20:58:53 +0100, Luis Gustavo wrote:
>> I've checked the following in.
>
> Still getting now these regressions in gdbserver non-extended mode:
>
> -PASS: gdb.base/save-bp.exp: info break
> +FAIL: gdb.base/save-bp.exp: info break
> +FAIL: gdb.cp/mb-templates.exp: initial condition: run to breakpoint (timeout)
> +FAIL: gdb.cp/mb-templates.exp: initial condition: run to breakpoint 2 (GDB internal error)
> +FAIL: gdb.cp/mb-templates.exp: ...

Yes. I am aware of those.

I've mentioned the mb-templates.exp one in the first mail about the 
series. It seems to be a problem that i uncovered when dealing with the 
agent expressions machinery on GDB's side. I'll debug this next.

The save-bp.exp one is due to the additional information that is output 
by info break. I'll have a patch for that soon (to make it accept 
additional info after the condition expression).

Luis

>
> info break
> Num     Type           Disp Enb Address            What
> 2       breakpoint     keep y   0x0000000000400540 in break_me at ./gdb.base/save-bp.c:21
> 3       breakpoint     keep y   0x000000000040054a in main at ./gdb.base/save-bp.c:27
> 4       breakpoint     keep y   0x000000000040054f in main at ./gdb.base/save-bp.c:28 thread 1
>          stop only in thread 1
> 5       breakpoint     keep y   0x000000000040055d in main at ./gdb.base/save-bp.c:31
>          stop only if i == 1 (host evals)
> 6       breakpoint     keep y   0x000000000040056c in main at ./gdb.base/save-bp.c:33
>          silent
> (gdb) FAIL: gdb.base/save-bp.exp: info break
>
> (I did not check now what is expeted.)
>
> (gdb) continue
> Continuing.
> ax-gdb.c:560: internal-error: gen_fetch: bad type code
> A problem internal to GDB has been detected,
> further debugging may prove unreliable.
> Quit this debugging session? (y or n) Found breakpoint condition.
> FAIL: gdb.cp/mb-templates.exp: initial condition: run to breakpoint (timeout)
>
>
> Regards,
> Jan

Re: Regressions in gdbserver mode #2 [Re: General regressions in gdbserver mode]

[Jan Kratochvil]

On Sun, 26 Feb 2012 23:56:58 +0100, Luis Gustavo wrote:
> On 02/26/2012 06:54 PM, Jan Kratochvil wrote:
> > Still getting now these regressions in gdbserver non-extended mode:
> > 
> > -PASS: gdb.base/save-bp.exp: info break
> > +FAIL: gdb.base/save-bp.exp: info break
> > +FAIL: gdb.cp/mb-templates.exp: initial condition: run to breakpoint (timeout)
> > +FAIL: gdb.cp/mb-templates.exp: initial condition: run to breakpoint 2 (GDB internal error)
> > +FAIL: gdb.cp/mb-templates.exp: ...
> 
> Yes. I am aware of those.
> 
> I've mentioned the mb-templates.exp one in the first mail about the
> series. It seems to be a problem that i uncovered when dealing with
> the agent expressions machinery on GDB's side. I'll debug this next.

I see here now "I'll debug this next.", I forgot about that.  Posted it now
as KFAILs:
	[patch 1/2] testsuite: Code cleanup: mb-templates.exp
	http://sourceware.org/ml/gdb-patches/2012-02/msg00655.html
	[patch 2/2] KFAIL for mb-templates.exp PASS->FAIL regression
	http://sourceware.org/ml/gdb-patches/2012-02/msg00656.html

Jan

Re: Regressions in gdbserver mode #2 [Re: General regressions in gdbserver mode]

[Luis Gustavo]

[-- Attachment #1: Type: text/plain, Size: 1654 bytes --]

On 02/26/2012 06:54 PM, Jan Kratochvil wrote:
> On Sat, 25 Feb 2012 20:58:53 +0100, Luis Gustavo wrote:
>> I've checked the following in.
>
> Still getting now these regressions in gdbserver non-extended mode:
>
> -PASS: gdb.base/save-bp.exp: info break
> +FAIL: gdb.base/save-bp.exp: info break
> +FAIL: gdb.cp/mb-templates.exp: initial condition: run to breakpoint (timeout)
> +FAIL: gdb.cp/mb-templates.exp: initial condition: run to breakpoint 2 (GDB internal error)
> +FAIL: gdb.cp/mb-templates.exp: ...
>
> info break
> Num     Type           Disp Enb Address            What
> 2       breakpoint     keep y   0x0000000000400540 in break_me at ./gdb.base/save-bp.c:21
> 3       breakpoint     keep y   0x000000000040054a in main at ./gdb.base/save-bp.c:27
> 4       breakpoint     keep y   0x000000000040054f in main at ./gdb.base/save-bp.c:28 thread 1
>          stop only in thread 1
> 5       breakpoint     keep y   0x000000000040055d in main at ./gdb.base/save-bp.c:31
>          stop only if i == 1 (host evals)
> 6       breakpoint     keep y   0x000000000040056c in main at ./gdb.base/save-bp.c:33
>          silent
> (gdb) FAIL: gdb.base/save-bp.exp: info break
>
> (I did not check now what is expeted.)
>
> (gdb) continue
> Continuing.
> ax-gdb.c:560: internal-error: gen_fetch: bad type code
> A problem internal to GDB has been detected,
> further debugging may prove unreliable.
> Quit this debugging session? (y or n) Found breakpoint condition.
> FAIL: gdb.cp/mb-templates.exp: initial condition: run to breakpoint (timeout)
>
>
> Regards,
> Jan

What about the following to fix save-bp.exp? Hopefully the syntax is 
correct.

Luis

[-- Attachment #2: save-bp-fix.diff --]
[-- Type: text/x-patch, Size: 1278 bytes --]

2012-02-27  Luis Machado  <lgustavo@codesourcery.com>

	* gdb.base/save-bp.exp: Account for new condition evaluator output.

diff --git a/gdb/testsuite/gdb.base/save-bp.exp b/gdb/testsuite/gdb.base/save-bp.exp
index 439a20e..7d63dc3 100644
--- a/gdb/testsuite/gdb.base/save-bp.exp
+++ b/gdb/testsuite/gdb.base/save-bp.exp
@@ -66,4 +66,4 @@ gdb_test "source bps"
 # Now, verify that all breakpoints have been created correctly...
 set bp_row_start "\[0-9\]+ +breakpoint +keep +y +0x\[0-9a-f\]+ +in"
 gdb_test "info break" \
-  " *Num +Type +Disp +Enb +Address +What\r\n$bp_row_start break_me at .*$srcfile:\[0-9\]+\r\n$bp_row_start main at .*$srcfile:$loc_bp2\r\n$bp_row_start main at .*$srcfile:$loc_bp3 +thread 1\r\n\[ \t]+stop only in thread 1\r\n$bp_row_start main at .*$srcfile:$loc_bp4\r\n\[ \t\]+stop only if i == 1\r\n$bp_row_start main at .*$srcfile:$loc_bp5\r\n\[ \t\]+silent"
+  " *Num +Type +Disp +Enb +Address +What\r\n$bp_row_start break_me at .*$srcfile:\[0-9\]+\r\n$bp_row_start main at .*$srcfile:$loc_bp2\r\n$bp_row_start main at .*$srcfile:$loc_bp3 +thread 1\r\n\[ \t]+stop only in thread 1\r\n$bp_row_start main at .*$srcfile:$loc_bp4\r\n\[ \t\]+stop only if i == 1( \\((host|target) evals\\))?\r\n$bp_row_start main at .*$srcfile:$loc_bp5\r\n\[ \t\]+silent"

Re: Regressions in gdbserver mode #2 [Re: General regressions in gdbserver mode]

[Jan Kratochvil]

On Mon, 27 Feb 2012 04:55:00 +0100, Luis Gustavo wrote:
> What about the following to fix save-bp.exp? Hopefully the syntax is
> correct.

> 2012-02-27  Luis Machado  <lgustavo@codesourcery.com>
> 
> 	* gdb.base/save-bp.exp: Account for new condition evaluator output.

Yes, I find the patch OK.


Thanks,
Jan

Re: Regressions in gdbserver mode #2 [Re: General regressions in gdbserver mode]

[Luis Gustavo]

On 02/27/2012 01:18 AM, Jan Kratochvil wrote:
> On Mon, 27 Feb 2012 04:55:00 +0100, Luis Gustavo wrote:
>> What about the following to fix save-bp.exp? Hopefully the syntax is
>> correct.
>
>> 2012-02-27  Luis Machado<lgustavo@codesourcery.com>
>>
>> 	* gdb.base/save-bp.exp: Account for new condition evaluator output.
>
> Yes, I find the patch OK.
>
>
> Thanks,
> Jan

Thanks, i've checked it in now.

Luis