Re: Octave-maintainers Digest, Vol 84, Issue 21

[Rik]

On 03/09/2013 09:10 AM, address@hidden wrote:
> Message: 4
> Date: Sat, 09 Mar 2013 00:44:33 -0600
> From: Daniel J Sebald <address@hidden>
> To: Sander van Rijn <address@hidden>
> Cc: address@hidden
> Subject: Re: Desired behavior for bug #38485: using 'run history'
>       twice
> Message-ID: <address@hidden>
> Content-Type: text/plain; charset=ISO-8859-1; format=flowed
>
> On 03/08/2013 09:48 PM, Daniel J Sebald wrote:
>> > On 03/08/2013 07:16 PM, Sander van Rijn wrote:
>>> >>
>>> >> Fri 08 Mar 2013 06:00:09 AM CET, *original submission:*
>>> >> <https://savannah.gnu.org/bugs/?38485#comment0>
>>> >>
>>> >> Using 'run history' twice in a row leads to a segfault. The first
>>> >> invocation runs the previous command, whatever it was. The second
>>> >> invocation executes the previous command which is 'run history' and an
>>> >> infinite loop is generated. This isn't a common occurrence, but it isn't
>>> >> great behavior by the interpreter either.
>>> >>
>>> >> Sample Code:
>>> >>
>>> >> disp (1); run history run history
>>> >>
>>> >>
>>> >>
>>> >> What would be the desired behavior in this case? An error about causing
>>> >> an infinite loop or a recursive implementation which would make second
>>> >> 'run history' in the sample code also execute 'disp (1);' ?
>> >
>> > I can't replicate this problem. What is "run history" supposed to do?
> Oh, run_history, with an underscore.  Yes, I see there is a problem 
> there.  I'm inclined to say recursive implementation, but could that run 
> into stack issues thereby requiring some error checking also or leave it 
> open to potentially crash?  Have to think this one over.
>
> I point out that the help documentation for "run_history" could be 
> improved as well.  Anyway, there may be a bug in the run_history whereby 
> the most recent history entry can't be accessed by number:
>
> octave:24> history -5
>   1048 1
>   1049 2
>   1050 3
>   1051 4
>   1052 history -5
> octave:25> run_history 1052
> error: run_history: history specification out of range
>
> The second most recent can be accessed though:
>
> octave:29> history -5
>   1054 6
>   1055 7
>   1056 8
>   1057 9
>   1058 history -5
> octave:30> run_history 1057
> ans =  9
>
> Lastly, the non-argument version of "history" doesn't seem of much use. 
>   Is the following what it is supposed to do?
>
> octave:34> history
>   1063 history
> octave:35> history
>   1064 history
>
> I would think that a full listing of history would be a more useful 
> thing to display, or at least default the list to, say, 25 entries.
>
> Dan
3/9/13

Dan,

Which Mercurial changeset are you running?  I improved both the
documentation and fixed the off-by-1 error, and it was while coding there
that I noticed this issue.

history with no arguments used to return all history starting from command
#1.  I think that is a regression although I didn't touch that part of the
code.  I'll take a look though.

--Rik