1. Variables address (Reformulated)
Sorry everyone,
I found a message from Robert Craig that repudiate the idea of
getting a euphoria variable address, as he say:
"I firmly believe this would be a very bad thing to do"
So, considering that, I want to reformulate my question.
Can anyone give me an idea of an efficient method to create something
like the ex.err file in a C translated program?
My current solution is to set a global variable with the
values of some relevant variables. I do that in many parts of the code
where I "feel" that a crash may occur, let´s put in another way...
any critical block in the code is preceded by a call to a routine that
set my global variable with some values passed as parameters,
usually variables used in the "critical block".
So, when a crash occur, another routine, designated by crash_routine(),
takes control and just prints out the values from my global variable.
What about that? Watch out SUN J2EE... just kidding,
I know it is a "very" amateur approach.
So, any creative ideas.
Thank you all,
André Drummond
2. Re: Variables address (Reformulated)
André Drummond wrote:
>
> Sorry everyone,
>
> I found a message from Robert Craig that repudiate the idea of
> getting a euphoria variable address, as he say:
>
> "I firmly believe this would be a very bad thing to do"
>
> So, considering that, I want to reformulate my question.
>
> Can anyone give me an idea of an efficient method to create something
> like the ex.err file in a C translated program?
>
> My current solution is to set a global variable with the
> values of some relevant variables. I do that in many parts of the code
> where I "feel" that a crash may occur, let´s put in another way...
> any critical block in the code is preceded by a call to a routine that
> set my global variable with some values passed as parameters,
> usually variables used in the "critical block".
>
> So, when a crash occur, another routine, designated by crash_routine(),
> takes control and just prints out the values from my global variable.
>
> What about that? Watch out SUN J2EE... just kidding,
> I know it is a "very" amateur approach.
>
> So, any creative ideas.
>
> Thank you all,
> André Drummond
>
Hello André,
I usually pass by answering this general kind of thing because I always imagine
others have more sophisticated techniques. However, I'll dive in with an idea:
what about not waiting for a crash to record what your program is doing? When I
am developing a program I set the constant DEBUG=1, and pepper my code with some
version of the code below:
if DEBUG then
state=AppendFile(DEBUGFILE,relevantData,"Current Prg Report-Section X")
CheckState(state,"DEBUG:File problem: AppendFile() failed on Report X")
end if
The constant DEBUGFILE is set to the fullpath of the debug file and AppendFile()
does what it says.
Obviously if the program fails at a certain point, the previously saved info is
readily available. If you use a program similar to Edita, you would just click
the debug file's tab and re-open it. And if some sections are working fine, just
delete or comment-out the DEBUG stuff. If you have no access to setting DEBUG,
then set up a way to turn debugging on and off.
--Quark
3. Re: Variables address (Reformulated)
> I found a message from Robert Craig that repudiate the idea of
> getting a euphoria variable address, as he say:
>
> "I firmly believe this would be a very bad thing to do"
I once hacked a dll function to return the address of a Euphoria
object passed to it. I lost the code, but I may be able to recreate
it. The problem is that you have to use an external source to "tell"
you what the address is, I don't really think there's a way to do it
from the interpreter.
The reason Rob thinks its a bad idea, I believe, is because of
reference counting, and by altering the variable via its address in
memory, you could cause a memory leak.
~Greg
