bug#64226: 30.0.50; emacs-lisp-native-compile-and-load permission error

bug#64226: 30.0.50; emacs-lisp-native-compile-and-load permission error

[No Wayman]

To reproduce:

1. Find an elisp file located in $DIR which user has permission to 
write to.
(in the following backtrace, the file being compiled was located 
at $HOME/.emacs.d/elpaca/builds/elpaca/elpaca.el)
2. M-x emacs-lisp-native-compile-and-load 
3. Observer the following error: 

Error: permission-denied (\"Creating file with prefix\" 
\"Permission denied\" 
\"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/elpaca-0646d6fc-1256ece5\")
  mapbacktrace(#f(compiled-function (evald func args flags) 
  #<bytecode 0x835cf36bbae8692>))
  debug-early-backtrace()
  debug-early(error (permission-denied \"Creating file with 
  prefix\" \"Permission denied\" 
  \"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/elpaca-0646d6fc-1256ece5\"))
  make-temp-file(\"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/elpaca-0646d6fc-1256ece5\" 
  nil \".eln.tmp\" nil)
  comp--compile-ctxt-to-file(\"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/elpaca-0646d6fc-1256ece5.eln\")
  comp-compile-ctxt-to-file(\"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/elpaca-0646d6fc-1256ece5.eln\")
  comp-final1()
  load-with-code-conversion(\"/tmp/emacs-int-comp-elpaca-0646d6fc-1256ece5-73Dpd0.el\" 
  \"/tmp/emacs-int-comp-elpaca-0646d6fc-1256ece5-73Dpd0.el\" nil 
  t)
  command-line-1((\"-l\" 
  \"/tmp/emacs-int-comp-elpaca-0646d6fc-1256ece5-73Dpd0.el\"))
  command-line()
  normal-top-level()

I would expect compiling this to not signal an error for a file I 
have permission to write.

In GNU Emacs 30.0.50 (build 1, x86_64-pc-linux-gnu, GTK+ Version
 3.24.38, cairo version 1.17.8) of 2023-06-19 built on laptop
Repository revision: edb0862f5e69240de90c30b8914af51778f26d31
Repository branch: master
Windowing system distributor 'The X.Org Foundation', version 
11.0.12101008
System Description: Arch Linux

I rebuilt from master and am still able to reproduce with
GNU Emacs 30.0.50 (build 1, x86_64-pc-linux-gnu, GTK+ Version 
3.24.38, cairo version 1.17.8) of 2023-06-21

bug#64226: 30.0.50; emacs-lisp-native-compile-and-load permission error

[Eli Zaretskii]

> From: No Wayman <iarchivedmywholelife@gmail.com>
> Date: Wed, 21 Jun 2023 12:58:40 -0400
> 
> 
> To reproduce:
> 
> 1. Find an elisp file located in $DIR which user has permission to 
> write to.
> (in the following backtrace, the file being compiled was located 
> at $HOME/.emacs.d/elpaca/builds/elpaca/elpaca.el)
> 2. M-x emacs-lisp-native-compile-and-load 
> 3. Observer the following error: 
> 
> Error: permission-denied (\"Creating file with prefix\" 
> \"Permission denied\" 
> \"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/elpaca-0646d6fc-1256ece5\")
>   mapbacktrace(#f(compiled-function (evald func args flags) 
>   #<bytecode 0x835cf36bbae8692>))
>   debug-early-backtrace()
>   debug-early(error (permission-denied \"Creating file with 
>   prefix\" \"Permission denied\" 
>   \"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/elpaca-0646d6fc-1256ece5\"))
>   make-temp-file(\"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/elpaca-0646d6fc-1256ece5\" 
>   nil \".eln.tmp\" nil)
>   comp--compile-ctxt-to-file(\"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/elpaca-0646d6fc-1256ece5.eln\")
>   comp-compile-ctxt-to-file(\"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/elpaca-0646d6fc-1256ece5.eln\")
>   comp-final1()
>   load-with-code-conversion(\"/tmp/emacs-int-comp-elpaca-0646d6fc-1256ece5-73Dpd0.el\" 
>   \"/tmp/emacs-int-comp-elpaca-0646d6fc-1256ece5-73Dpd0.el\" nil 
>   t)
>   command-line-1((\"-l\" 
>   \"/tmp/emacs-int-comp-elpaca-0646d6fc-1256ece5-73Dpd0.el\"))
>   command-line()
>   normal-top-level()
> 
> I would expect compiling this to not signal an error for a file I 
> have permission to write.

Is this in "emacs -Q"?  If not, please see if "emacs -Q" reproduces
the problem.

I also don't understand why the backtrace above seems to imply that
you invoked Emacs like this:

  emacs -l /tmp/emacs-int-comp-elpaca-0646d6fc-1256ece5-73Dpd0.el

IOW, the backtrace doesn't show invocation of
emacs-lisp-native-compile-and-load, it shows the attempt to load an
already-compiled .eln file.  What am I missing?  I added Andrea to
this discussion.

If, in "emacs -Q", you visit the source file, that is

  $HOME/.emacs.d/elpaca/builds/elpaca/elpaca.el

and then type

  M-: (native-compile buffer-file-name) RET

what file name do you see in the echo-area after the compilation
finishes?

bug#64226: 30.0.50; emacs-lisp-native-compile-and-load permission error

[No Wayman]

Eli Zaretskii <eliz@gnu.org> writes:

>> From: No Wayman <iarchivedmywholelife@gmail.com>
>> Date: Wed, 21 Jun 2023 12:58:40 -0400
>> 
>> 
>> To reproduce:
>> 
>> 1. Find an elisp file located in $DIR which user has permission 
>> to 
>> write to.
>> (in the following backtrace, the file being compiled was 
>> located 
>> at $HOME/.emacs.d/elpaca/builds/elpaca/elpaca.el)
>> 2. M-x emacs-lisp-native-compile-and-load 
>> 3. Observer the following error: 
>> 
>> Error: permission-denied (\"Creating file with prefix\" 
>> \"Permission denied\" 
>> \"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/elpaca-0646d6fc-1256ece5\")
>>   mapbacktrace(#f(compiled-function (evald func args flags) 
>>   #<bytecode 0x835cf36bbae8692>))
>>   debug-early-backtrace()
>>   debug-early(error (permission-denied \"Creating file with 
>>   prefix\" \"Permission denied\" 
>>   \"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/elpaca-0646d6fc-1256ece5\"))
>>   make-temp-file(\"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/elpaca-0646d6fc-1256ece5\"
>>   nil \".eln.tmp\" nil)
>>   comp--compile-ctxt-to-file(\"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/elpaca-0646d6fc-1256ece5.eln\")
>>   comp-compile-ctxt-to-file(\"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/elpaca-0646d6fc-1256ece5.eln\")
>>   comp-final1()
>>   load-with-code-conversion(\"/tmp/emacs-int-comp-elpaca-0646d6fc-1256ece5-73Dpd0.el\"
>>   \"/tmp/emacs-int-comp-elpaca-0646d6fc-1256ece5-73Dpd0.el\" 
>>   nil 
>>   t)
>>   command-line-1((\"-l\" 
>>   \"/tmp/emacs-int-comp-elpaca-0646d6fc-1256ece5-73Dpd0.el\"))
>>   command-line()
>>   normal-top-level()
>> 
>> I would expect compiling this to not signal an error for a file 
>> I 
>> have permission to write.
 
> Is this in "emacs -Q"?  If not, please see if "emacs -Q" 
> reproduces
> the problem.

I have reproduced from emacs -Q as well. Same error as above:


Compiling /home/n/.emacs.d/elpaca/repos/elpaca/elpaca.el...done
comp--native-compile: Native compiler error: 
"/home/n/.emacs.d/elpaca/repos/elpaca/elpaca.el", "Compiling 
/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/elpaca-0646d6fc-1256ece5.eln...
Creating file with prefix: Permission denied, 
/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/elpaca-0646d6fc-1256ece5

Error: permission-denied (\"Creating file with prefix\" 
\"Permission denied\" 
\"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/elpaca-0646d6fc-1256ece5\")
  mapbacktrace(#f(compiled-function (evald func args flags) 
  #<bytecode 0x70d6f07f3ae8645>))
  debug-early-backtrace()
  debug-early(error (permission-denied \"Creating file with 
  prefix\" \"Permission denied\" 
  \"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/elpaca-0646d6fc-1256ece5\"))
  make-temp-file(\"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/elpaca-0646d6fc-1256ece5\" 
  nil \".eln.tmp\" nil)
  comp--compile-ctxt-to-file(\"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/elpaca-0646d6fc-1256ece5.eln\")
  comp-compile-ctxt-to-file(\"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/elpaca-0646d6fc-1256ece5.eln\")
  comp-final1()
  load-with-code-conversion(\"/tmp/emacs-int-comp-elpaca-0646d6fc-1256ece5-oSdhiB.el\" 
  \"/tmp/emacs-int-comp-elpaca-0646d6fc-1256ece5-oSdhiB.el\" nil 
  t)
  command-line-1((\"-l\" 
  \"/tmp/emacs-int-comp-elpaca-0646d6fc-1256ece5-oSdhiB.el\"))
  command-line()
  normal-top-level()
 
> I also don't understand why the backtrace above seems to imply 
> that
> you invoked Emacs like this:
>
>   emacs -l 
>   /tmp/emacs-int-comp-elpaca-0646d6fc-1256ece5-73Dpd0.el

> IOW, the backtrace doesn't show invocation of
> emacs-lisp-native-compile-and-load, it shows the attempt to load 
> an
> already-compiled .eln file.  What am I missing?  I added Andrea 
> to
> this discussion.

The compilation takes place in a subprocess, no?
That's probably how the subprocess is invoked.
 

bug#64226: 30.0.50; emacs-lisp-native-compile-and-load permission error

[Eli Zaretskii]

> From: No Wayman <iarchivedmywholelife@gmail.com>
> Cc: Andrea Corallo <acorallo@gnu.org>, 64226@debbugs.gnu.org
> Date: Thu, 22 Jun 2023 11:44:33 -0400
> 
> I have reproduced from emacs -Q as well. Same error as above:
> 
> 
> Compiling /home/n/.emacs.d/elpaca/repos/elpaca/elpaca.el...done
> comp--native-compile: Native compiler error: 
> "/home/n/.emacs.d/elpaca/repos/elpaca/elpaca.el", "Compiling 
> /usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/elpaca-0646d6fc-1256ece5.eln...
> Creating file with prefix: Permission denied, 
> /usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/elpaca-0646d6fc-1256ece5
> 
> Error: permission-denied (\"Creating file with prefix\" 
> \"Permission denied\" 
> \"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/elpaca-0646d6fc-1256ece5\")
>   mapbacktrace(#f(compiled-function (evald func args flags) 
>   #<bytecode 0x70d6f07f3ae8645>))
>   debug-early-backtrace()
>   debug-early(error (permission-denied \"Creating file with 
>   prefix\" \"Permission denied\" 
>   \"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/elpaca-0646d6fc-1256ece5\"))
>   make-temp-file(\"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/elpaca-0646d6fc-1256ece5\" 
>   nil \".eln.tmp\" nil)
>   comp--compile-ctxt-to-file(\"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/elpaca-0646d6fc-1256ece5.eln\")
>   comp-compile-ctxt-to-file(\"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/elpaca-0646d6fc-1256ece5.eln\")
>   comp-final1()
>   load-with-code-conversion(\"/tmp/emacs-int-comp-elpaca-0646d6fc-1256ece5-oSdhiB.el\" 
>   \"/tmp/emacs-int-comp-elpaca-0646d6fc-1256ece5-oSdhiB.el\" nil 
>   t)
>   command-line-1((\"-l\" 
>   \"/tmp/emacs-int-comp-elpaca-0646d6fc-1256ece5-oSdhiB.el\"))
>   command-line()
>   normal-top-level()
>  
> > I also don't understand why the backtrace above seems to imply 
> > that
> > you invoked Emacs like this:
> >
> >   emacs -l 
> >   /tmp/emacs-int-comp-elpaca-0646d6fc-1256ece5-73Dpd0.el
> 
> > IOW, the backtrace doesn't show invocation of
> > emacs-lisp-native-compile-and-load, it shows the attempt to load 
> > an
> > already-compiled .eln file.  What am I missing?  I added Andrea 
> > to
> > this discussion.
> 
> The compilation takes place in a subprocess, no?

No, I don't think so, not with emacs-lisp-native-compile-and-load.
Andrea, am I right?

And you haven't answered my other question:

> If, in "emacs -Q", you visit the source file, that is
> 
>   $HOME/.emacs.d/elpaca/builds/elpaca/elpaca.el
> 
> and then type
> 
>   M-: (native-compile buffer-file-name) RET
> 
> what file name do you see in the echo-area after the compilation
> finishes?

bug#64226: 30.0.50; emacs-lisp-native-compile-and-load permission error

[No Wayman]

Eli Zaretskii <eliz@gnu.org> writes:

> No, I don't think so, not with 
> emacs-lisp-native-compile-and-load.
> Andrea, am I right?

See comp-final:

	(with-temp-buffer
          (unwind-protect
              (if (zerop
                   (call-process (expand-file-name invocation-name
                                                   invocation-directory)
				 nil t t "-no-comp-spawn" "-Q" "--batch" "-l"
                                 temp-file))
                  (progn
                    (delete-file temp-file)
                    output)
		(signal 'native-compiler-error (list (buffer-string))))
            (comp-log-to-buffer (buffer-string))))))))

 
> And you haven't answered my other question:
 
>> If, in "emacs -Q", you visit the source file, that is
>> 
>>   $HOME/.emacs.d/elpaca/builds/elpaca/elpaca.el
>> 
>> and then type
>> 
>>   M-: (native-compile buffer-file-name) RET
>> 
>> what file name do you see in the echo-area after the 
>> compilation
>> finishes?

The following file name is messaged with no error signaled:

"/home/n/.emacs.d/eln-cache/30.0.50-58a66af6/elpaca-0646d6fc-1256ece5.eln"

bug#64226: 30.0.50; emacs-lisp-native-compile-and-load permission error

[Eli Zaretskii]

> From: No Wayman <iarchivedmywholelife@gmail.com>
> Cc: acorallo@gnu.org, 64226@debbugs.gnu.org
> Date: Thu, 22 Jun 2023 12:22:07 -0400
> 
> 
> Eli Zaretskii <eliz@gnu.org> writes:
> 
> > No, I don't think so, not with 
> > emacs-lisp-native-compile-and-load.
> > Andrea, am I right?
> 
> See comp-final:

That's not compilation, that's code generation, the last stage of
compilation.

> >> If, in "emacs -Q", you visit the source file, that is
> >> 
> >>   $HOME/.emacs.d/elpaca/builds/elpaca/elpaca.el
> >> 
> >> and then type
> >> 
> >>   M-: (native-compile buffer-file-name) RET
> >> 
> >> what file name do you see in the echo-area after the 
> >> compilation
> >> finishes?
> 
> The following file name is messaged with no error signaled:
> 
> "/home/n/.emacs.d/eln-cache/30.0.50-58a66af6/elpaca-0646d6fc-1256ece5.eln"

Which is exactly what emacs-lisp-native-compile-and-load does.  So I
wonder what went wrong in your case.

What is the value of native-comp-eln-load-path?

Also, can you show the contents of the file being compiled, which
seems to be /home/n/.emacs.d/elpaca/repos/elpaca/elpaca.el ?

bug#64226: 30.0.50; emacs-lisp-native-compile-and-load permission error

[No Wayman]

Eli Zaretskii <eliz@gnu.org> writes:
 
> What is the value of native-comp-eln-load-path?

("/home/n/.emacs.d/eln-cache/" 
"/usr/lib/emacs/30.0.50/native-lisp/")

 
> Also, can you show the contents of the file being compiled, 
> which
> seems to be /home/n/.emacs.d/elpaca/repos/elpaca/elpaca.el ?

I can reproduce with any elisp file.
e.g. $HOME/test.el containing:

--8<---------------cut here---------------start------------->8---
;; -*- lexical-binding: t; -*-
(defun +test () (message "PASS"))
--8<---------------cut here---------------end--------------->8---


Native compiler error: "/home/n/test.el", "Compiling 
/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/test-98b414c9-ac29ecab.eln...
Creating file with prefix: Permission denied, 
/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/test-98b414c9-ac29ecab

Error: permission-denied (\"Creating file with prefix\" 
\"Permission denied\" 
\"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/test-98b414c9-ac29ecab\")
  mapbacktrace(#f(compiled-function (evald func args flags) 
  #<bytecode 0xd45c107f3ae87e6>))
  debug-early-backtrace()
  debug-early(error (permission-denied \"Creating file with 
  prefix\" \"Permission denied\" 
  \"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/test-98b414c9-ac29ecab\"))
  make-temp-file(\"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/test-98b414c9-ac29ecab\" 
  nil \".eln.tmp\" nil)
  comp--compile-ctxt-to-file(\"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/test-98b414c9-ac29ecab.eln\")
  comp-compile-ctxt-to-file(\"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/test-98b414c9-ac29ecab.eln\")
  comp-final1()
  load-with-code-conversion(\"/tmp/emacs-int-comp-test-98b414c9-ac29ecab-my5A7o.el\" 
  \"/tmp/emacs-int-comp-test-98b414c9-ac29ecab-my5A7o.el\" nil t)
  command-line-1((\"-l\" 
  \"/tmp/emacs-int-comp-test-98b414c9-ac29ecab-my5A7o.el\"))
  command-line()
  normal-top-level()

bug#64226: 30.0.50; emacs-lisp-native-compile-and-load permission error

[Eli Zaretskii]

> From: No Wayman <iarchivedmywholelife@gmail.com>
> Cc: acorallo@gnu.org, 64226@debbugs.gnu.org
> Date: Thu, 22 Jun 2023 13:21:49 -0400
> 
> 
> Eli Zaretskii <eliz@gnu.org> writes:
>  
> > What is the value of native-comp-eln-load-path?
> 
> ("/home/n/.emacs.d/eln-cache/" 
> "/usr/lib/emacs/30.0.50/native-lisp/")
> 
>  
> > Also, can you show the contents of the file being compiled, 
> > which
> > seems to be /home/n/.emacs.d/elpaca/repos/elpaca/elpaca.el ?
> 
> I can reproduce with any elisp file.
> e.g. $HOME/test.el containing:
> 
> --8<---------------cut here---------------start------------->8---
> ;; -*- lexical-binding: t; -*-
> (defun +test () (message "PASS"))
> --8<---------------cut here---------------end--------------->8---

Then I'm stumped, and hope Andrea will be able to figure this out.  I
cannot reproduce the problem on my system, FWIW.

bug#64226: 30.0.50; emacs-lisp-native-compile-and-load permission error

[Andrea Corallo]

Eli Zaretskii <eliz@gnu.org> writes:

>> From: No Wayman <iarchivedmywholelife@gmail.com>
>> Cc: acorallo@gnu.org, 64226@debbugs.gnu.org
>> Date: Thu, 22 Jun 2023 13:21:49 -0400
>> 
>> 
>> Eli Zaretskii <eliz@gnu.org> writes:
>>  
>> > What is the value of native-comp-eln-load-path?
>> 
>> ("/home/n/.emacs.d/eln-cache/" 
>> "/usr/lib/emacs/30.0.50/native-lisp/")
>> 
>>  
>> > Also, can you show the contents of the file being compiled, 
>> > which
>> > seems to be /home/n/.emacs.d/elpaca/repos/elpaca/elpaca.el ?
>> 
>> I can reproduce with any elisp file.
>> e.g. $HOME/test.el containing:
>> 
>> --8<---------------cut here---------------start------------->8---
>> ;; -*- lexical-binding: t; -*-
>> (defun +test () (message "PASS"))
>> --8<---------------cut here---------------end--------------->8---
>
> Then I'm stumped, and hope Andrea will be able to figure this out.  I
> cannot reproduce the problem on my system, FWIW.

That's very bizzare to me as well, I can't reproduce this either.  Is
anyone able to reproduce this?

Thanks

  Andrea

bug#64226: 30.0.50; emacs-lisp-native-compile-and-load permission error

[Eli Zaretskii]

Ping!  Can anyone reproduce this issue and report the details, please?

> From: Andrea Corallo <acorallo@gnu.org>
> Cc: No Wayman <iarchivedmywholelife@gmail.com>,  64226@debbugs.gnu.org
> Date: Sun, 25 Jun 2023 11:02:41 -0400
> 
> Eli Zaretskii <eliz@gnu.org> writes:
> 
> >> From: No Wayman <iarchivedmywholelife@gmail.com>
> >> Cc: acorallo@gnu.org, 64226@debbugs.gnu.org
> >> Date: Thu, 22 Jun 2023 13:21:49 -0400
> >> 
> >> 
> >> Eli Zaretskii <eliz@gnu.org> writes:
> >>  
> >> > What is the value of native-comp-eln-load-path?
> >> 
> >> ("/home/n/.emacs.d/eln-cache/" 
> >> "/usr/lib/emacs/30.0.50/native-lisp/")
> >> 
> >>  
> >> > Also, can you show the contents of the file being compiled, 
> >> > which
> >> > seems to be /home/n/.emacs.d/elpaca/repos/elpaca/elpaca.el ?
> >> 
> >> I can reproduce with any elisp file.
> >> e.g. $HOME/test.el containing:
> >> 
> >> --8<---------------cut here---------------start------------->8---
> >> ;; -*- lexical-binding: t; -*-
> >> (defun +test () (message "PASS"))
> >> --8<---------------cut here---------------end--------------->8---
> >
> > Then I'm stumped, and hope Andrea will be able to figure this out.  I
> > cannot reproduce the problem on my system, FWIW.
> 
> That's very bizzare to me as well, I can't reproduce this either.  Is
> anyone able to reproduce this?
> 
> Thanks
> 
>   Andrea
> 

bug#64226: 30.0.50; emacs-lisp-native-compile-and-load permission error

[No Wayman]

Eli Zaretskii <eliz@gnu.org> writes:

> Ping!  Can anyone reproduce this issue and report the details, 
> please?

Still reproducible on my end on two machines:

1. GNU Emacs 30.0.50 (build 1, x86_64-pc-linux-gnu, GTK+ Version 
3.24.38, cairo version 1.17.8) of 2023-06-28

2. GNU Emacs 30.0.50 (build 1, x86_64-pc-linux-gnu, GTK+ Version 
3.24.38, cairo version 1.17.8) of 2023-06-15

Native compiler error: 
"/home/n/.emacs.d/elpaca/repos/elpaca/elpaca.el", "Compiling 
/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/elpaca-0646d6fc-bfea2b95.eln...
Creating file with prefix: Permission denied, 
/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/elpaca-0646d6fc-bfea2b95

Error: permission-denied (\"Creating file with prefix\" 
\"Permission denied\" 
\"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/elpaca-0646d6fc-bfea2b95\")
  mapbacktrace(#f(compiled-function (evald func args flags) 
  #<bytecode 0x1e734bf67ae8647>))
  debug-early-backtrace()
  debug-early(error (permission-denied \"Creating file with 
  prefix\" \"Permission denied\" 
  \"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/elpaca-0646d6fc-bfea2b95\"))
  make-temp-file(\"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/elpaca-0646d6fc-bfea2b95\" 
  nil \".eln.tmp\" nil)
  comp--compile-ctxt-to-file(\"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/elpaca-0646d6fc-bfea2b95.eln\")
  comp-compile-ctxt-to-file(\"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/elpaca-0646d6fc-bfea2b95.eln\")
  comp-final1()
  load-with-code-conversion(\"/tmp/emacs-int-comp-elpaca-0646d6fc-bfea2b95-SXTWTL.el\" 
  \"/tmp/emacs-int-comp-elpaca-0646d6fc-bfea2b95-SXTWTL.el\" nil 
  t)
  command-line-1((\"-l\" 
  \"/tmp/emacs-int-comp-elpaca-0646d6fc-bfea2b95-SXTWTL.el\"))
  command-line()
  normal-top-level()
"

bug#64226: 30.0.50; emacs-lisp-native-compile-and-load permission error

[Eli Zaretskii]

> From: No Wayman <iarchivedmywholelife@gmail.com>
> Cc: Andrea Corallo <acorallo@gnu.org>, 64226@debbugs.gnu.org
> Date: Sat, 01 Jul 2023 11:28:49 -0400
> 
> 
> Eli Zaretskii <eliz@gnu.org> writes:
> 
> > Ping!  Can anyone reproduce this issue and report the details, 
> > please?
> 
> Still reproducible on my end on two machines:

Thanks, but that's a small wonder: we didn't yet do anything to fix
this, whatever it is.

I was asking others to provide details, not just reproduce.  If you
want to provide those details, perhaps you could describe how to
reproduce starting from "emacs -Q" (and loading any packages or
features you need for the minimal reproduction recipe).  The way you
were showing the problem until now obviously depends on your local
setup, which is impossible to reproduce without knowing the details.

The main aspect of this which is completely unclear is: how come Emacs
tries to write in the /usr/lib tree when compiling Lisp files from
your home directory.  This is not supposed to happen: Emacs should
write to the eln-cache subdirectory of your ~/.emacs.d/ directory.

bug#64226: 30.0.50; emacs-lisp-native-compile-and-load permission error

[No Wayman]

Eli Zaretskii <eliz@gnu.org> writes:

> perhaps you could describe how to
> reproduce starting from "emacs -Q" (and loading any packages or
> features you need for the minimal reproduction recipe).  The way 
> you
> were showing the problem until now obviously depends on your 
> local
> setup, which is impossible to reproduce without knowing the 
> details.

The instructions I gave here:

https://lists.gnu.org/archive/html/bug-gnu-emacs/2023-06/msg01224.html

reproduce the error on my end from emacs -Q with a minimal file.
To be crystal clear:

1. Save the following file to $HOME/test.el (or any path where the 
user has write permissions):

--8<---------------cut here---------------start------------->8---
;; -*- lexical-binding: t; -*-
(defun +test () (message "PASS"))
--8<---------------cut here---------------end--------------->8---

2. emacs -Q $HOME/test.el
3. M-x emacs-lisp-native-compile-compile-and-load
4. Error is printed to message buffer.

I'm not sure how I can meaningfully pare it down from there.

> The main aspect of this which is completely unclear is: how come 
> Emacs
> tries to write in the /usr/lib tree when compiling Lisp files 
> from
> your home directory.  This is not supposed to happen: Emacs 
> should
> write to the eln-cache subdirectory of your ~/.emacs.d/ 
> directory.

That makes sense to me and I agree.
It looks like the critical path is in the file 
`comp-spill-lap-function' method.
The comp-ctxt-output slot is set to (car (last 
native-comp-eln-load-path)).

I don't know enough about the native-comp machinery to say whether 
or not this is the appropriate solution, but let-binding 
`native-compile-target-directory' to the car of 
native-comp-eln-load-path in emacs-lisp-native-compile-and-load 
compiles in the correct directory and avoids the error:

(defun emacs-lisp-native-compile-and-load ()
  "Native-compile synchronously the current file (if it has 
  changed).
Load the compiled code when finished.

Use `emacs-lisp-byte-compile-and-load' in combination with
`native-comp-jit-compilation' set to t to achieve asynchronous
native compilation."
  (interactive nil emacs-lisp-mode)
  (emacs-lisp--before-compile-buffer)
  (let ((byte+native-compile t)
        (native-compile-target-directory (car 
        native-comp-eln-load-path))
        (byte-to-native-output-buffer-file nil))
    (when-let ((eln (native-compile buffer-file-name)))
      (load (file-name-sans-extension (comp-write-bytecode-file 
      eln))))))

I'll leave the work of a proper patch to someone who is more 
familiar with the system.

bug#64226: 30.0.50; emacs-lisp-native-compile-and-load permission error

[Eli Zaretskii]

> From: No Wayman <iarchivedmywholelife@gmail.com>
> Cc: acorallo@gnu.org, 64226@debbugs.gnu.org
> Date: Sat, 01 Jul 2023 15:14:16 -0400
> 
> 
> Eli Zaretskii <eliz@gnu.org> writes:
> 
> > perhaps you could describe how to reproduce starting from "emacs
> > -Q" (and loading any packages or features you need for the minimal
> > reproduction recipe).  The way you were showing the problem until
> > now obviously depends on your local setup, which is impossible to
> > reproduce without knowing the details.
> 
> The instructions I gave here:
> 
> https://lists.gnu.org/archive/html/bug-gnu-emacs/2023-06/msg01224.html
> 
> reproduce the error on my end from emacs -Q with a minimal file.
> To be crystal clear:
> 
> 1. Save the following file to $HOME/test.el (or any path where the 
> user has write permissions):
> 
> --8<---------------cut here---------------start------------->8---
> ;; -*- lexical-binding: t; -*-
> (defun +test () (message "PASS"))
> --8<---------------cut here---------------end--------------->8---
> 
> 2. emacs -Q $HOME/test.el
> 3. M-x emacs-lisp-native-compile-compile-and-load
> 4. Error is printed to message buffer.

Thanks.  I think Andrea couldn't reproduce this, but maybe he could
try again.  (The error will only happen if the directory
usr/lib/emacs/30.0.50/native-lisp is not writable by your user, so
ensuring it's not writable should be part of the recipe.)

Also, please show the backtrace corresponding to the above recipe.
The backtrace you posted mentions files like
/home/n/.emacs.d/elpaca/repos/elpaca/elpaca.el, but there's no
reference to $HOME/elpaca/repos in your recipe above.  So I presume
the backtrace with the above recipe will mention different files, and
it is important to see those file names.

bug#64226: 30.0.50; emacs-lisp-native-compile-and-load permission error

[No Wayman]

Eli Zaretskii <eliz@gnu.org> writes:

> Also, please show the backtrace corresponding to the above 
> recipe.
> The backtrace you posted mentions files like
> /home/n/.emacs.d/elpaca/repos/elpaca/elpaca.el, but there's no
> reference to $HOME/elpaca/repos in your recipe above.  So I 
> presume
> the backtrace with the above recipe will mention different 
> files, and
> it is important to see those file names.

I sincerely doubt that will show anything more than what has been 
demonstrated already. 
Here you are from emacs -Q /tmp/test.el (contents mentioned in 
last message):

Compiling /tmp/test.el...done
comp--native-compile: Native compiler error: "/tmp/test.el", 
"Compiling 
/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/test-53c62054-ac29ecab.eln...
Creating file with prefix: Permission denied, 
/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/test-53c62054-ac29ecab

Error: permission-denied (\"Creating file with prefix\" 
\"Permission denied\" 
\"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/test-53c62054-ac29ecab\")
  mapbacktrace(#f(compiled-function (evald func args flags) 
  #<bytecode -0x6de363c98517970>))
  debug-early-backtrace()
  debug-early(error (permission-denied \"Creating file with 
  prefix\" \"Permission denied\" 
  \"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/test-53c62054-ac29ecab\"))
  make-temp-file(\"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/test-53c62054-ac29ecab\" 
  nil \".eln.tmp\" nil)
  comp--compile-ctxt-to-file(\"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/test-53c62054-ac29ecab.eln\")
  comp-compile-ctxt-to-file(\"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/test-53c62054-ac29ecab.eln\")
  comp-final1()
  load-with-code-conversion(\"/tmp/emacs-int-comp-test-53c62054-ac29ecab-s3wF7W.el\" 
  \"/tmp/emacs-int-comp-test-53c62054-ac29ecab-s3wF7W.el\" nil t)
  command-line-1((\"-l\" 
  \"/tmp/emacs-int-comp-test-53c62054-ac29ecab-s3wF7W.el\"))
  command-line()
  normal-top-level()
"

bug#64226: 30.0.50; emacs-lisp-native-compile-and-load permission error

[Andrea Corallo]

Eli Zaretskii <eliz@gnu.org> writes:

>> From: No Wayman <iarchivedmywholelife@gmail.com>
>> Cc: acorallo@gnu.org, 64226@debbugs.gnu.org
>> Date: Sat, 01 Jul 2023 15:14:16 -0400
>> 
>> 
>> Eli Zaretskii <eliz@gnu.org> writes:
>> 
>> > perhaps you could describe how to reproduce starting from "emacs
>> > -Q" (and loading any packages or features you need for the minimal
>> > reproduction recipe).  The way you were showing the problem until
>> > now obviously depends on your local setup, which is impossible to
>> > reproduce without knowing the details.
>> 
>> The instructions I gave here:
>> 
>> https://lists.gnu.org/archive/html/bug-gnu-emacs/2023-06/msg01224.html
>> 
>> reproduce the error on my end from emacs -Q with a minimal file.
>> To be crystal clear:
>> 
>> 1. Save the following file to $HOME/test.el (or any path where the 
>> user has write permissions):
>> 
>> --8<---------------cut here---------------start------------->8---
>> ;; -*- lexical-binding: t; -*-
>> (defun +test () (message "PASS"))
>> --8<---------------cut here---------------end--------------->8---
>> 
>> 2. emacs -Q $HOME/test.el
>> 3. M-x emacs-lisp-native-compile-compile-and-load
>> 4. Error is printed to message buffer.
>
> Thanks.  I think Andrea couldn't reproduce this, but maybe he could
> try again.  (The error will only happen if the directory
> usr/lib/emacs/30.0.50/native-lisp is not writable by your user, so
> ensuring it's not writable should be part of the recipe.)

Just retried and can finally reproduce the error.  Will look at.

Thanks

  Andrea

bug#64226: 30.0.50; emacs-lisp-native-compile-and-load permission error

[Andrea Corallo]

Andrea Corallo <acorallo@gnu.org> writes:

> Eli Zaretskii <eliz@gnu.org> writes:
>
>>> From: No Wayman <iarchivedmywholelife@gmail.com>
>>> Cc: acorallo@gnu.org, 64226@debbugs.gnu.org
>>> Date: Sat, 01 Jul 2023 15:14:16 -0400
>>> 
>>> 
>>> Eli Zaretskii <eliz@gnu.org> writes:
>>> 
>>> > perhaps you could describe how to reproduce starting from "emacs
>>> > -Q" (and loading any packages or features you need for the minimal
>>> > reproduction recipe).  The way you were showing the problem until
>>> > now obviously depends on your local setup, which is impossible to
>>> > reproduce without knowing the details.
>>> 
>>> The instructions I gave here:
>>> 
>>> https://lists.gnu.org/archive/html/bug-gnu-emacs/2023-06/msg01224.html
>>> 
>>> reproduce the error on my end from emacs -Q with a minimal file.
>>> To be crystal clear:
>>> 
>>> 1. Save the following file to $HOME/test.el (or any path where the 
>>> user has write permissions):
>>> 
>>> --8<---------------cut here---------------start------------->8---
>>> ;; -*- lexical-binding: t; -*-
>>> (defun +test () (message "PASS"))
>>> --8<---------------cut here---------------end--------------->8---
>>> 
>>> 2. emacs -Q $HOME/test.el
>>> 3. M-x emacs-lisp-native-compile-compile-and-load
>>> 4. Error is printed to message buffer.
>>
>> Thanks.  I think Andrea couldn't reproduce this, but maybe he could
>> try again.  (The error will only happen if the directory
>> usr/lib/emacs/30.0.50/native-lisp is not writable by your user, so
>> ensuring it's not writable should be part of the recipe.)
>
> Just retried and can finally reproduce the error.  Will look at.
>
> Thanks
>
>   Andrea

Okay I pushed b93107c20b2 to fix this.  I tested as best as I could but
tweaking these dynamic variables is always tricky.

No Wayman if you could verify everything works for you would be great.

Thanks

  Andrea

bug#64226: 30.0.50; emacs-lisp-native-compile-and-load permission error

[No Wayman]

Andrea Corallo <acorallo@gnu.org> writes:

> Okay I pushed b93107c20b2 to fix this.  I tested as best as I 
> could but
> tweaking these dynamic variables is always tricky.
>
> No Wayman if you could verify everything works for you would be 
> great.
>
> Thanks
>
>   Andrea

Still present on:

GNU Emacs 30.0.50 (build 1, x86_64-pc-linux-gnu, GTK+ Version 
3.24.38, cairo version 1.17.8) of 2023-08-09


Fixed on:

GNU Emacs 29.1.50 (build 3, x86_64-pc-linux-gnu, GTK+ Version 
3.24.38, cairo version 1.17.8) of 2023-08-09


Thank you.

bug#64226: 30.0.50; emacs-lisp-native-compile-and-load permission error

[Eli Zaretskii]

> From: No Wayman <iarchivedmywholelife@gmail.com>
> Cc: Eli Zaretskii <eliz@gnu.org>, 64226@debbugs.gnu.org
> Date: Wed, 09 Aug 2023 11:24:03 -0400
> 
> 
> Andrea Corallo <acorallo@gnu.org> writes:
> 
> > Okay I pushed b93107c20b2 to fix this.  I tested as best as I 
> > could but
> > tweaking these dynamic variables is always tricky.
> >
> > No Wayman if you could verify everything works for you would be 
> > great.
> >
> > Thanks
> >
> >   Andrea
> 
> Still present on:
> 
> GNU Emacs 30.0.50 (build 1, x86_64-pc-linux-gnu, GTK+ Version 
> 3.24.38, cairo version 1.17.8) of 2023-08-09

That's expected, since the fix was installed on the emacs-29 branch,
and will be merged to master only in a couple of days.

Thanks for testing (and thanks to Andrea for fixing it), I'm therefore
closing this bug.

bug#64226: 30.0.50; emacs-lisp-native-compile-and-load permission error

[Eli Zaretskii]

> From: Andrea Corallo <acorallo@gnu.org>
> Cc: No Wayman <iarchivedmywholelife@gmail.com>,  64226@debbugs.gnu.org
> Date: Wed, 09 Aug 2023 09:22:35 -0400
> 
> Okay I pushed b93107c20b2 to fix this.  I tested as best as I could but
> tweaking these dynamic variables is always tricky.

These changes broke building Emacs in the source tree: now all the
*.eln files produced by the build's ELC+ELN rule are written to the
user's eln-cache directory instead of into native-lisp subdirectory of
the build tree.  To reproduce, touch, say, lisp/files.el or delete
lisp/files.elc, and then say "make" -- you will see that the *.eln
file is written under ~/.emacs.d/eln-cache, and if you also delete the
corresponding files-NNNN.eln file from native-lisp, Emacs will load
files.elc during loadup.

bug#64226: 30.0.50; emacs-lisp-native-compile-and-load permission error

[Andrea Corallo]

Eli Zaretskii <eliz@gnu.org> writes:

>> From: Andrea Corallo <acorallo@gnu.org>
>> Cc: No Wayman <iarchivedmywholelife@gmail.com>,  64226@debbugs.gnu.org
>> Date: Wed, 09 Aug 2023 09:22:35 -0400
>> 
>> Okay I pushed b93107c20b2 to fix this.  I tested as best as I could but
>> tweaking these dynamic variables is always tricky.
>
> These changes broke building Emacs in the source tree: now all the
> *.eln files produced by the build's ELC+ELN rule are written to the
> user's eln-cache directory instead of into native-lisp subdirectory of
> the build tree.  To reproduce, touch, say, lisp/files.el or delete
> lisp/files.elc, and then say "make" -- you will see that the *.eln
> file is written under ~/.emacs.d/eln-cache, and if you also delete the
> corresponding files-NNNN.eln file from native-lisp, Emacs will load
> files.elc during loadup.

Thanks, looking at it.

  Andrea

bug#64226: 30.0.50; emacs-lisp-native-compile-and-load permission error

[Andrea Corallo]

Andrea Corallo <acorallo@gnu.org> writes:

> Eli Zaretskii <eliz@gnu.org> writes:
>
>>> From: Andrea Corallo <acorallo@gnu.org>
>>> Cc: No Wayman <iarchivedmywholelife@gmail.com>,  64226@debbugs.gnu.org
>>> Date: Wed, 09 Aug 2023 09:22:35 -0400
>>> 
>>> Okay I pushed b93107c20b2 to fix this.  I tested as best as I could but
>>> tweaking these dynamic variables is always tricky.
>>
>> These changes broke building Emacs in the source tree: now all the
>> *.eln files produced by the build's ELC+ELN rule are written to the
>> user's eln-cache directory instead of into native-lisp subdirectory of
>> the build tree.  To reproduce, touch, say, lisp/files.el or delete
>> lisp/files.elc, and then say "make" -- you will see that the *.eln
>> file is written under ~/.emacs.d/eln-cache, and if you also delete the
>> corresponding files-NNNN.eln file from native-lisp, Emacs will load
>> files.elc during loadup.
>
> Thanks, looking at it.

Okay, just pushed 842dbf500e0, seems to solve the issue here for me,
please have a look.

Thanks!

  Andrea

bug#64226: 30.0.50; emacs-lisp-native-compile-and-load permission error

[Eli Zaretskii]

> From: Andrea Corallo <acorallo@gnu.org>
> Cc: iarchivedmywholelife@gmail.com,  64226@debbugs.gnu.org
> Date: Sat, 12 Aug 2023 12:56:38 -0400
> 
> Andrea Corallo <acorallo@gnu.org> writes:
> 
> > Eli Zaretskii <eliz@gnu.org> writes:
> >
> >>> From: Andrea Corallo <acorallo@gnu.org>
> >>> Cc: No Wayman <iarchivedmywholelife@gmail.com>,  64226@debbugs.gnu.org
> >>> Date: Wed, 09 Aug 2023 09:22:35 -0400
> >>> 
> >>> Okay I pushed b93107c20b2 to fix this.  I tested as best as I could but
> >>> tweaking these dynamic variables is always tricky.
> >>
> >> These changes broke building Emacs in the source tree: now all the
> >> *.eln files produced by the build's ELC+ELN rule are written to the
> >> user's eln-cache directory instead of into native-lisp subdirectory of
> >> the build tree.  To reproduce, touch, say, lisp/files.el or delete
> >> lisp/files.elc, and then say "make" -- you will see that the *.eln
> >> file is written under ~/.emacs.d/eln-cache, and if you also delete the
> >> corresponding files-NNNN.eln file from native-lisp, Emacs will load
> >> files.elc during loadup.
> >
> > Thanks, looking at it.
> 
> Okay, just pushed 842dbf500e0, seems to solve the issue here for me,
> please have a look.

Thanks, works correctly now.

bug#64226: 30.0.50; emacs-lisp-native-compile-and-load permission error

[Andrea Corallo]

Eli Zaretskii <eliz@gnu.org> writes:

>> From: No Wayman <iarchivedmywholelife@gmail.com>
>> Cc: Andrea Corallo <acorallo@gnu.org>, 64226@debbugs.gnu.org
>> Date: Thu, 22 Jun 2023 11:44:33 -0400
>> 
>> I have reproduced from emacs -Q as well. Same error as above:
>> 
>> 
>> Compiling /home/n/.emacs.d/elpaca/repos/elpaca/elpaca.el...done
>> comp--native-compile: Native compiler error: 
>> "/home/n/.emacs.d/elpaca/repos/elpaca/elpaca.el", "Compiling 
>> /usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/elpaca-0646d6fc-1256ece5.eln...
>> Creating file with prefix: Permission denied, 
>> /usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/elpaca-0646d6fc-1256ece5
>> 
>> Error: permission-denied (\"Creating file with prefix\" 
>> \"Permission denied\" 
>> \"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/elpaca-0646d6fc-1256ece5\")
>>   mapbacktrace(#f(compiled-function (evald func args flags) 
>>   #<bytecode 0x70d6f07f3ae8645>))
>>   debug-early-backtrace()
>>   debug-early(error (permission-denied \"Creating file with 
>>   prefix\" \"Permission denied\" 
>>   \"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/elpaca-0646d6fc-1256ece5\"))
>>   make-temp-file(\"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/elpaca-0646d6fc-1256ece5\" 
>>   nil \".eln.tmp\" nil)
>>   comp--compile-ctxt-to-file(\"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/elpaca-0646d6fc-1256ece5.eln\")
>>   comp-compile-ctxt-to-file(\"/usr/lib/emacs/30.0.50/native-lisp/30.0.50-58a66af6/elpaca-0646d6fc-1256ece5.eln\")
>>   comp-final1()
>>   load-with-code-conversion(\"/tmp/emacs-int-comp-elpaca-0646d6fc-1256ece5-oSdhiB.el\" 
>>   \"/tmp/emacs-int-comp-elpaca-0646d6fc-1256ece5-oSdhiB.el\" nil 
>>   t)
>>   command-line-1((\"-l\" 
>>   \"/tmp/emacs-int-comp-elpaca-0646d6fc-1256ece5-oSdhiB.el\"))
>>   command-line()
>>   normal-top-level()
>>  
>> > I also don't understand why the backtrace above seems to imply 
>> > that
>> > you invoked Emacs like this:
>> >
>> >   emacs -l 
>> >   /tmp/emacs-int-comp-elpaca-0646d6fc-1256ece5-73Dpd0.el
>> 
>> > IOW, the backtrace doesn't show invocation of
>> > emacs-lisp-native-compile-and-load, it shows the attempt to load 
>> > an
>> > already-compiled .eln file.  What am I missing?  I added Andrea 
>> > to
>> > this discussion.
>> 
>> The compilation takes place in a subprocess, no?
>
> No, I don't think so, not with emacs-lisp-native-compile-and-load.
> Andrea, am I right?

We always invoke libgccjit in a subprocess, the difference between sync
and async is:

- in sync compilation we run all our compiler *but* libgccjit in process
  (libgccjit runs in a subprocess)

- in async all the compilation computation is done in a sub process.

The reason for running libgccjit always in a subprocess is that
unfortunatelly it leaks memory.

Best Regards

  Andrea