Re: [reverse RFC] Add documentation for process record and replay

[teawater <teawater at gmail dot com>]

Hi Eli,

On Wed, Oct 15, 2008 at 00:33, Eli Zaretskii <eliz@gnu.org> wrote:
>> Date: Tue, 14 Oct 2008 23:00:59 +0800
>> From: teawater <teawater@gmail.com>
>>
>> This patch add doc to gdb.texinfo.
>>
>> 2008-10-14  Hui Zhu  <teawater@gmail.com>
>>
>>       * gdb.texinfo: Add documentation for process record and replay.
>
> Thank you.  I have a few questions and suggestions about this.
>

Thanks for your questions and suggestions. All of them are very good.

>> +When this target is in use, if the next instruction to be executed is in the
>> +execution log, @value{GDBN} will debug in replay mode so that all the
>> +execution events are taken from the execution log.  Otherwise, @value{GDBN}
>> +will debug in record mode and record the execution log while executing
>> +normally.
>
> What exactly is the "execution log"?  You talk about "next
> instruction", which seems to hint that the log records the machine
> instructions executed by the inferior -- is that true?  If so, what
> are the "execution events" you mention here -- are they just a synonym
> for the "instructions", or are they something else?
>

"execution log" mean is before a instruction execute, record the
values of register and memory that will be change in this instruction
to a list. This list is called "execution log".
In replay mode, inferior will not execute actually. It will just
execute in GDB itself. So this called "execution events".

I am trying to talk clear this in a internal doc that I am writing.


>> +Process record and replay target can only debug a process that already
>> +running. Therefore you need to first start the process @code{run},
>> +and then start the recording @code{record}.
>
> This is really inconvenient, because it means the beginning of the
> program's execution cannot be recorded.  Is that a limitation, or am I
> missing something?

Cause P record will record the values before instruction execute. So
it can't work before program's begin.
And I don't think we need it. Cause most of time, people just care
about program execute after "main". You can set a breakpoint in "main"
and open P record when inferior break in there.
Or if you care about a lot of thing before "main". You can set a
breakpoint in "_start" and open P record when inferior break in there.


>
> This:
>
>> +@item stoprecord
>> +Stop process record and replay target at once. When Process record and
>> +replay target stops, all the execution log will be deleted and the inferior
>> +will either be terminated, or remain in its final state.
>
> and this:
>
>> +When you stop the process record and replay target in record mode (at the
>> +end of the execution log), the inferior will be stopped at the next
>> +instruction that would have been recorded.
>
> seem to be in contradiction: the former says that the inferior will be
> terminated, the latter says that it will be stopped.  Which one is it?

"terminated" is a suggest from Michael. I think it's mean is stop in there.

>
>>                                              In other words, if you record
>> +for a while and then stop recording, the inferior process will be left in
>> +the same state as if you had never done any recording.
>
> Why is this sentence important?  Recording has no side effects except
> for the recorded information, so why did you need to say that the
> inferior will be left in a state ``as if recording never happened''?
>

OK. I will change it.



>> +@itemx set record-auto-delete 1
>> +Set the behavior when record instructions limit is reached. 1 mean that GDB
>> +will auto delete the oldest record to make room for each new one.
>> +
>> +@item set record-auto-delete 0
>> +0 is the default value, meaning that @value{GDBN} will stop the inferior.
>
> Either the name of the option or the argument should be modified,
> IMO.  I prefer to rename the option (to something like
> record-stop-at-limit, and change the default to 1).
>
> Also, there's a relation between this option and
> record-insn-number-max, right?  If so, the text should mention that.
>

I think 0 is better than 1. Cause when GDB stop, it will ask user how
to do stop or set record-auto-delete to 1. I think it's user
experience is better than default to 1. Maybe I need add more
introduce to  record-auto-delete 0.

"record-stop-at-limit" is better. I will change it. :)



>> +@item set record-insn-number-max @var{limit}
>> +Set the limit of instructions to be recorded.  Default value is 200000.
>> +If set to 0, record instructions number limit function will disable.
>> +In this case, if record instructions number is bigger than @var{limit},
>> +@value{GDBN} will auto delete the earliest recorded instruction to make room
>> +for each new one.
>
> I think you need to transpose the two last sentences, otherwise the
> text seems to say that when the limit is set to zero, GDB will delete
> earliest instruction when the (zero) limit is exhausted.
>

I will transpose the two last sentences.

If the limit is set to zero, it will delete nothing. I will add some
introduce to this part.


Thanks,
Hui