native-comp build failures

native-comp build failures

[Alex Bennée]

Hi,

I've started tracking the native-comp feature branch and the results so
far have been impressive. However the build broke for me today so I was
wondering if there where any pointers on how to fix things?

  Dumping under the name bootstrap-emacs.pdmp
  dumping fingerprint: aa918a1f070b5959481eb75cab1ab2138d096c3661fe882973c9d197fda8ed9c
  Dump complete
  Byte counts: header=96 hot=8075448 discardable=138224 cold=5307672
  Reloc counts: hot=443393 discardable=4819
  make -C ../lisp compile-first EMACS="../src/bootstrap-emacs"
  make[2]: Entering directory '/home/alex/src/emacs/emacs.git/lisp'
   ELC+ELN   emacs-lisp/bytecomp.elc
   ELC+ELN   emacs-lisp/comp.elc
  Package subdirs is deprecated
  Package subdirs is deprecated
  Compiler-macro error for cl--block-wrapper: (wrong-number-of-arguments #<subr cl--block-wrapper--cmacro> 2)
  Compiler-macro error for cl--block-wrapper: (wrong-number-of-arguments #<subr cl--block-wrapper--cmacro> 2)

  ... repeats a lot ...
  
  Compiler-macro error for cl--block-wrapper: (wrong-number-of-arguments #<subr cl--block-wrapper--cmacro> 2)
  Compiler-macro error for cl--block-wrapper: (wrong-number-of-arguments #<subr cl--block-wrapper--cmacro> 2)
  Wrong number of arguments: #<subr cl--block-wrapper--cmacro>, 2
  make[2]: *** [Makefile:312: emacs-lisp/comp.elc] Error 255
  make[2]: *** Waiting for unfinished jobs....
  Compiler-macro error for cl--block-wrapper: (wrong-number-of-arguments #<subr cl--block-wrapper--cmacro> 2)
  Compiler-macro error for cl--block-wrapper: (wrong-number-of-arguments #<subr cl--block-wrapper--cmacro> 2)
  Wrong type argument: listp, #<subr byte-compile-form>
  make[2]: *** [Makefile:312: emacs-lisp/bytecomp.elc] Error 255
  make[2]: Leaving directory '/home/alex/src/emacs/emacs.git/lisp'
  make[1]: *** [Makefile:827: bootstrap-emacs.pdmp] Error 2
  make[1]: Leaving directory '/home/alex/src/emacs/emacs.git/src'
  make: *** [Makefile:424: src] Error 2


I've also notices that while GNUS seems to run faster when I start it I
see the message:

   You should byte-compile Gnus

So is there a way to tell which functions are byte-compiled and which
have been compiled into native code?

-- 
Alex Bennée

Re: native-comp build failures

[Andrea Corallo]

Alex Bennée <alex.bennee@linaro.org> writes:

> Hi,
>
> I've started tracking the native-comp feature branch and the results so
> far have been impressive. However the build broke for me today so I was
> wondering if there where any pointers on how to fix things?
>

Hi Alex,

thanks for trying this out.

Just a question, was the build folder clean in the git sense?  Consider
make clean is currently broken (I want get to it this weekend).
https://debbugs.gnu.org/cgi/bugreport.cgi?bug=39985

I ask that because I've boostraped it on my dev machine and built the
docker, also the CI is also clean:

https://emba.gnu.org/emacs/emacs/pipelines/5014

Bests

  Andrea

-- 
akrl@sdf.org

Re: native-comp build failures

[Alex Bennée]

Andrea Corallo <akrl@sdf.org> writes:

> Alex Bennée <alex.bennee@linaro.org> writes:
>
>> Hi,
>>
>> I've started tracking the native-comp feature branch and the results so
>> far have been impressive. However the build broke for me today so I was
>> wondering if there where any pointers on how to fix things?
>>
>
> Hi Alex,
>
> thanks for trying this out.
>
> Just a question, was the build folder clean in the git sense?  Consider
> make clean is currently broken (I want get to it this weekend).
> https://debbugs.gnu.org/cgi/bugreport.cgi?bug=39985

I did a:

  make distclean
  ./autogen.sh
  ./configure --with-x-toolkit=lucid --prefix=/home/alex/src/emacs/install --with-modules --with-imagemagick --with-nativecomp
  make

but it appears a rm -rf *; and git reset --hard HEAD works!

>
> I ask that because I've boostraped it on my dev machine and built the
> docker, also the CI is also clean:
>
> https://emba.gnu.org/emacs/emacs/pipelines/5014
>
> Bests
>
>   Andrea


-- 
Alex Bennée

Re: native-comp build failures

[Andrea Corallo]

Sorry I managed to miss the second part of the mail.

Alex Bennée <alex.bennee@linaro.org> writes:
> I've also notices that while GNUS seems to run faster when I start it I
> see the message:
>
>    You should byte-compile Gnus

Yeah I see the same.  I guess this is because somewhere gnus is checking
for a function to be bytecompiled and assume that if it's not it must be
interpreted.

> So is there a way to tell which functions are byte-compiled and which
> have been compiled into native code?

A part from `describe-function' programmatically you can do:

(type-of (symbol-function 'org-mode)) => compiled-function

becomes when native compiled:

(type-of (symbol-function 'org-mode)) => subr

you can use predicated `subrp' and `subr-native-elisp-p' too.

Andrea

--
akrl@sdf.org