New trace option?

[Ann and Jason Edmeades]

I just noticed there is an option --debugmsg +tid which adds the thread
prefix to the debugmsg trace. I would really, really like a timestamp
prefix - Is this something people would/could find acceptable?

I've hacked a version for my own use using gettimeofday, but anything which
shows seconds or slightly higher resolution would be invaluable. Perhaps a
+time option?

Is this something people would find useful? If so, anyone know a good,
portable way of doing it?

Jason

[Andreas Mohr]

On Sat, Apr 19, 2003 at 05:22:01PM +0100, Ann and Jason Edmeades wrote:
> I just noticed there is an option --debugmsg +tid which adds the thread
> prefix to the debugmsg trace. I would really, really like a timestamp
> prefix - Is this something people would/could find acceptable?
>
> I've hacked a version for my own use using gettimeofday, but anything which
> shows seconds or slightly higher resolution would be invaluable. Perhaps a
> +time option?
>
> Is this something people would find useful? If so, anyone know a good,
> portable way of doing it?

I'd find this very useful, of course especially for profiling purposes.
I don't have any specific idea how to implement it, though.

[Eric Pouech]

> I'd find this very useful, of course especially for profiling purposes.
> I don't have any specific idea how to implement it, though.

on P4s (and above), you could use RDTSC for an efficient and high 
resolution timestamp

A+

[Ferenc Wagner]

"Ann and Jason Edmeades"  writes:
> I would really, really like a timestamp prefix

\begin{wishes}
Yes, that would be usable.  And I would also very much like
a way to switch debugging on/off temporarily, since it makes
wine very slow even when not needed.  And I would also like
a lot of other fancy things, like searching, filtering,
matching, indenting etc. during the run.
\end{wishes}

\begin{idea}
All of the above could be decoupled of Wine if buffering
did not spoil interactivity when using pipes, and if pipe
buffers were possible to overflow.  There is an expect
script called unbuffer to solve the first issue.  Once I
tried to put together a program which forwards/drops lines
according to SIGUSR sent to him.  I failed.  Less or Emacs
could do the rest.  Use Perl for timestamping in the middle.
\end{idea}

Perhaps I will do this one day.  I am still not sure if the
design is right.  Seems generally useful, I wonder it is not
done yet...
                                             Feri.