advice.el and bytecomp: Did something change recently?

advice.el and bytecomp: Did something change recently?

[T.V Raman]

Context:

1. This is with emacs compiled from git @HEAD.
2. Emacspeak uses advice heavily (advice.el, not nadvice).
3. Until recently, advising functions that were loaded later still
worked,
4. EG:  Module emacspeak-advice.el advices vc-next-action and gets
loaded  before  vc.el.
5. Later, when vc.el gets loaded as a result of calling
vc-next-action, the advice would take effect automatically.

Now:

Since the last few days, (noticed it first yesterday  I think) the
above is no longer true.
I need to explicitly load vc.el and then emacspeak-advice.el  for the
advice on vc-next-action to take effect.

-- 

-- 

Re: advice.el and bytecomp: Did something change recently?

[Stefan Monnier]

> Since the last few days, (noticed it first yesterday  I think) the
> above is no longer true.
> I need to explicitly load vc.el and then emacspeak-advice.el  for the
> advice on vc-next-action to take effect.

That would be a bug (both for `defadvice` and for `advice-add`, and
currently this is handled by nadvice.el in both cases).


        Stefan

Re: advice.el and bytecomp: Did something change recently?

[T.V Raman]

Thanks for the quick response  Stefan!

Is there any additional information that I can provide --- warning:
The emacspeak usage is likely a more complex case than what may be
needed to debug this.
Stefan Monnier writes:
 > > Since the last few days, (noticed it first yesterday  I think) the
 > > above is no longer true.
 > > I need to explicitly load vc.el and then emacspeak-advice.el  for the
 > > advice on vc-next-action to take effect.
 > 
 > That would be a bug (both for `defadvice` and for `advice-add`, and
 > currently this is handled by nadvice.el in both cases).
 > 
 > 
 >         Stefan

-- 
Id: kg:/m/0285kf1 

-- 
Id: kg:/m/0285kf1 

Re: advice.el and bytecomp: Did something change recently?

[Stefan Monnier]

> Is there any additional information that I can provide

A reproduceable test case?
When I try

    M-: (defadvice sm-foo (before test activate) (message "before-advice")) RET
    M-: (defun sm-foo () (message "body")) RET
    M-: (sm-foo) RET

the advice works just fine.


        Stefan

Re: advice.el and bytecomp: Did something change recently?

[T.V Raman]

this is what I meant by "it's hard to debug with the complexity of
Emacspeak".

I've failed to produce a minimal repro case --- what I do observe:

1. Some advice  defs are not taking effect i.e. I dont get the
expected auditory feedback.

2. Evaluating the advice definition makes it work, even though the
file containing the advice was already loaded when it was failing.
Stefan Monnier writes:
 > > Is there any additional information that I can provide
 > 
 > A reproduceable test case?
 > When I try
 > 
 >     M-: (defadvice sm-foo (before test activate) (message "before-advice")) RET
 >     M-: (defun sm-foo () (message "body")) RET
 >     M-: (sm-foo) RET
 > 
 > the advice works just fine.
 > 
 > 
 >         Stefan

-- 
Id: kg:/m/0285kf1 

-- 
Id: kg:/m/0285kf1 

Re: advice.el and bytecomp: Did something change recently?

[T.V Raman]

p.s. perhaps byte compilation has something to do with it --- all of
emacspeak's modules are bytecompiled.
Stefan Monnier writes:
 > > Is there any additional information that I can provide
 > 
 > A reproduceable test case?
 > When I try
 > 
 >     M-: (defadvice sm-foo (before test activate) (message "before-advice")) RET
 >     M-: (defun sm-foo () (message "body")) RET
 >     M-: (sm-foo) RET
 > 
 > the advice works just fine.
 > 
 > 
 >         Stefan

-- 
Id: kg:/m/0285kf1 

-- 
Id: kg:/m/0285kf1 

Re: advice.el and bytecomp: Did something change recently?

[T.V Raman]

Hi Stefan,

After recompiling *everything* within reach: i.e.

1. make dist clean in emacs
2. git pull -a
3 Complete clean build
4. Then compiled all the emacspeak files using the freshly built emacs

The problem went away.

I still have no idea what the issue was, 


--