[edbrowse-dev] a crude 'step'

[Kevin Carhart <kevin@carhart.net>]

I sort of innovated something while working on Maersk.  You can step a 
short sequence of code without having so many breakpoints that it's 
impractical.

First you add a flag to the top of startwindow just for this purpose, like 
stepper = false

Now you turn trace on and study the trace counter numbers a bit.  You have 
to identify where you want the breakpoints to start, which would be 
whichever counter number you would be setting an eval($bp) on, if using it 
as it works currently.

Now you have to write into trace, but not just turning alert3 into 
alert3;eval($bp)

You write in with 2 things.  The first involves both levels.  The trace 
routine, and the string where trace writes a+ alert +b.  In trace, you 
have a conditional.  It tests the counter number and says if the counter 
number has reached your desired start value (a1000) , you write in a 
string to a+alert+b.  This piece of text says, stepper= true.  For any 
other counter number, add nothing.

The second thing will be added to a+alert+b unconditionally.  It just says 
if stepper = true then eval($bp).

So this means that once a1000 is reached, it will continue to breakpoint 
at every possible juncture from then on, but you can also switch
stepping back off, by entering stepper=false at the breakpoint prompt!

So you can step the flow without needing to know what that flow is, and 
also be able to make it somewhat fine-grained.

I fear my ssh session is about to crash so I am sending here!



  is that you hardcode 
a conditional 
to say that if the counter number has reached a1000, stepper =true.  This 
happens at the time that the trace function is running because that is 
when the counter numbers are available.

The second thing is a level down.  It's JS that goes into the string that 
trace writes.