Hi all,
I currently have a use-case where I need to rely on `f_lasti`, but it seems
that it's not really possible to rely on it as it doesn't accurately point
to the actual bytecode last executed when PREDICT is used.
So, I'd like to know: is it possible to change that so that the `f_lasti`
is indeed properly updated (even when PREDICT) is used (to actually match
what the docs say about it: https://docs.python.org/3/library/inspect.html)?
As a note, my use case is doing a step into a function in a debugger.
This is needed in order to know whether a given function call maps the
target being step into.
For instance, given something as:
def method(c):
return c+1
a = method
a(a(a(1)))
The user can choose to step into the `a` at the middle. The only
conceivable way out I found to implement it reliably was by identifying
that when the user enters the `method()` function the f_lasti on the parent
frame is the one which would result in that function call.
So, I have this working on a POC, but to have it working on all the cases
I'm having to handle all existing bytecode instructions and then going to
ceval.c and then identify which ones have a PREDICT and then having to do
the adjustments to match the `f_lasti` accordingly when the function call
is reached (which is a bit of an herculean task to support multiple
versions of CPython as that's definitely an implementation detail), so, I'd
really be interested in having the f_lasti work reliably.
Thanks,
Fabio
I currently have a use-case where I need to rely on `f_lasti`, but it seems
that it's not really possible to rely on it as it doesn't accurately point
to the actual bytecode last executed when PREDICT is used.
So, I'd like to know: is it possible to change that so that the `f_lasti`
is indeed properly updated (even when PREDICT) is used (to actually match
what the docs say about it: https://docs.python.org/3/library/inspect.html)?
As a note, my use case is doing a step into a function in a debugger.
This is needed in order to know whether a given function call maps the
target being step into.
For instance, given something as:
def method(c):
return c+1
a = method
a(a(a(1)))
The user can choose to step into the `a` at the middle. The only
conceivable way out I found to implement it reliably was by identifying
that when the user enters the `method()` function the f_lasti on the parent
frame is the one which would result in that function call.
So, I have this working on a POC, but to have it working on all the cases
I'm having to handle all existing bytecode instructions and then going to
ceval.c and then identify which ones have a PREDICT and then having to do
the adjustments to match the `f_lasti` accordingly when the function call
is reached (which is a bit of an herculean task to support multiple
versions of CPython as that's definitely an implementation detail), so, I'd
really be interested in having the f_lasti work reliably.
Thanks,
Fabio