This is the mail archive of the gdb-patches@sourceware.org mailing list for the GDB project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH]tracepoint.c: Add conditionals num to find_matching_tracepoint


On Tue, Aug 9, 2011 at 07:02, Stan Shebs <stanshebs@earthlink.net> wrote:
> On 8/7/11 9:28 AM, Hui Zhu wrote:
>>
>> Hi,
>>
>> I found that when I set some tracepoint to a same address, and use
>> tsave. ? And use "target tfile " to open it. ?It will only one
>> tracepoint available.
>> And output some words like:
>> Created tracepoint 1 for target's tracepoint 1 at 0x40050a.
>> Assuming tracepoint 1 is same as target's tracepoint 2 at 0x40050a.
>> Assuming tracepoint 1 is same as target's tracepoint 3 at 0x40050a.
>>
>> This is because find_matching_tracepoint didn't check the num.
>
>
> The number is exactly the one property that tracepoint upload must never
> consider when looking for matching tracepoints, since the numbers vary
> depending on what the user has been doing during the current GDB session.
> ?Addressing the FIXME will help the multiple-tracepoint case, although it's
> kind of messy.
>
> There is another useful heuristic that would be easy to add, which is to
> exclude matching on a tracepoint that has already been uploaded. ?Having
> created tracepoint 1 from the uploaded info, it's never going to be the case
> that uploaded tracepoints 2 and 3 are the same as 1. ?It might be as simple
> as testing number_on_target, but beware that it might be nonzero due to
> other trace runs or some such.
>
> Stan
> stan@codesourcery.com
>
>>
>> And I think the tracepoint have the same address is really helpful for
>> user. ?Because we can set different condition to make tracepoint more
>> powerful.
>> So I make following patch.
>>
>> Please help me review it.
>>
>> Thanks,
>> Hui
>>
>>
>> 2011-08-08 ?Hui Zhu<teawater@gmail.com>
>>
>> ? ? ? ?* tracepoint.c (find_matching_tracepoint): Add number check.
>
>

Hi Stan,

Thanks for your review.

I make a new patch that check the condition according to your mail.

Best,
Hui

2011-08-14  Hui Zhu  <teawater@gmail.com>

	* tracepoint.c (cond_string_is_same): New function.
	(find_matching_tracepoint): Add condition check
	by cond_string_is_same.
---
 tracepoint.c |   19 ++++++++++++++++++-
 1 file changed, 18 insertions(+), 1 deletion(-)

--- a/tracepoint.c
+++ b/tracepoint.c
@@ -3091,6 +3091,22 @@ free_uploaded_tsvs (struct uploaded_tsv
     }
 }

+static int
+cond_string_is_same(char *str1, char *str2)
+{
+  if (str1 == NULL || str2 == NULL)
+    {
+      if (str1 == str2)
+	return 1;
+      else
+	return 0;
+    }
+  if (strcmp (str1, str2))
+    return 0;
+
+  return 1;
+}
+
 /* Look for an existing tracepoint that seems similar enough to the
    uploaded one.  Enablement isn't compared, because the user can
    toggle that freely, and may have done so in anticipation of the
@@ -3111,7 +3127,8 @@ find_matching_tracepoint (struct uploade
       if (b->type == utp->type
 	  && t->step_count == utp->step
 	  && t->pass_count == utp->pass
-	  /* FIXME also test conditionals and actions.  */
+	  && cond_string_is_same (t->base.cond_string, utp->cond_string)
+	  /* FIXME also test actions.  */
 	  )
 	{
 	  /* Scan the locations for an address match.  */


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]