Re: Native Compile: Example Of Possibly Spurious Warning

[T.V Raman]

Andrea Corallo <akrl@sdf.org> writes:


Hi Andrea,

emacspeak is at https://github.com/tvraman/emacspeak

I've tried many things, including batch native compiling emacspeak  ---
and to ensure that a running Emacs did not end up polluting the
eln-cache, I did this while running an emacs that did not have native
support compiled in.
If after then killing the non-native emacs and start the Emacs with
native support, I still see the warning I showed in my email.

Not sure whether you'll have the cycles to run emacspeak itself --
you'll minimally need to install the espeak and espeak-ng  packages to
get speech synthesis on Linux for Emacspeak to use.

Here is my suspicion:

When emacspeak is ormally compiled via make, it loads it's own loaddefs
file etc ---  and I suspect that when Emacs itself compiles any of the
emacspeak modules while running, it doesn't pull in the prerequisites.

Also: I was hoping that once batch-native-compile had done its thing for
all the emacspeak modules, then a newly started emacs would not need to
compile those files; however this is not true -- with no further changes
to the source-code of those files, Emacs appears to recompile those
files on the fly -- dont know why. I've not fully understood how the
cacheing strategy works, but I suspect it's over-aggressive.

One final thought: for complex packages like Emacspeak that have their
own custom build rules, it would be  nice if emacs' async native compile
cycle could be given access to those rules.




> Eli Zaretskii <eliz@gnu.org> writes:
>
>>> From: "T.V Raman" <raman@google.com>
>>> Cc: emacs-devel@gnu.org
>>> Date: Mon, 17 May 2021 06:58:17 -0700
>>> 
>>> Eli Zaretskii <eliz@gnu.org> writes:
>>> 
>>> 
>>> All of the spurious warnings I'm seeing dont appear in batch compilation
>>> with batch-byte-compile.
>>> Here is the file on Github:
>>> 
>>> https://github.com/tvraman/emacspeak/blob/master/lisp/emacspeak-preamble.el#L91
>>
>> I hope Andrea will look into this and tell what causes the error (my
>> guess is some other file this one loads or requires, but that's just a
>> guess).
>
> I'll be happy to have a look this week.
>
> Raman can you share a sequence precise instructions I can follow to
> reproduce what you see?
>
> Thanks
>
>   Andrea

-- 

Thanks,

--Raman
?7?4 Id: kg:/m/0285kf1  ?0?8