Re: Save database?

[Raymond Rogers <raymond rogers72 gmail com>]

Thank you for your attention; I thought I had enrolled but I guess not.

On 01/06/2015 05:38 PM, Dodji Seketeli wrote:

    I'm trying to debug some complicated math routines and need to
generate "trails" through various routines that can be checked and
compared by myself; a human.

Do you mean that you need to generate logs whenever control reaches
certain points of the program you are debugging?

Well I would rather do the logging across blocks; say start/stop line by line logging.

If that is what you want, then I am thinking that maybe you should use
SystemTap[1] for that.  It should allow you to write fairly sophisticated
scripts that let you monitor (and log) what happens during the execution
of the program you are interested in.

https://sourceware.org/systemtap/documentation.html

I will look but I think the mistake (or whatever it is) is mathematical in nature not programming.  That is, the wrong branches are taken when the same problem is stated different ways.  Which is easy to fix in theory but other data follows the same paths and has to take the "wrong" path so some delicacy is required in altering the program.

    Since session saves are already included is there some way to
translate the database into a English information?

Hmmh.  A Nemiver session is just a record of the context useful to debug
a program. That is, it contains things like the path to the program,
it's environment, the breakpoints you have set during the debugging and
things like that.  It doesn't contain anything about the internals of
the program being debugged that you would need to be able to emit traces
about where you have stepped into.  We could imagine adding a feature
like this in the future, though.

??  But you can resume a situation from the log including data values.  Therefore data values and program positions have to be in the database?

If not can somebody point me to somewhere in the code I could log the
instruction and data sequence.

Before I start pointing you to places int he code, I'd like to know if
you think that logging some context information each time the inferior
(the program being debugged) is stopped somewhere would be helpful to
you?  By context information I mean things like "where you have
stopped".  That can be file/line information of where you stopped at.
>From that I guess you could get the source-code-level statements you
stopped at, etc.  Would that be useful to your use case?

Yes I think so but start/stop line logging as blocks would be a lot more usefull.  The purpose is to coompare how I get to various points via the decision trees.  This might seem trivial; but I can assure you that it has not been for me.  The underlying problem is mathematical but that is reflected in the coding.  The justification and documentation is at best skimpy.

Cheers,

Have a good new year
Ray