bug#46502: 28.0.50; [feature/native-comp] (d3a399dd) native-comp bootstrap failure

[Pip Cet]

On Thu, Feb 18, 2021 at 10:04 AM Andrea Corallo via Bug reports for
GNU Emacs, the Swiss army knife of text editors
<bug-gnu-emacs@gnu.org> wrote:
> Michael Welsh Duggan <mwd@md5i.com> writes:
> > I decided to attempt building the native-comp feature branch, since it
> > seems like it may make it to mainline fairly soon.  When bootstrapping
> > with
> >
> >   make NATIVE_FULL_AOT=1 -j 32 bootstrap
> >
> > I get an error compiling gnus/gnus-icalendar.el.  I've put the log
> > below.  I've updated the emacs information from my running emacs below
> > to match the version of the repository from which I built.
> >
> >
> > In GNU Emacs 28.0.50 (build 1, x86_64-pc-linux-gnu, X toolkit, cairo 
> > version 1.16.0)
> >  of 2021-02-13 built on miko
> > Repository revision: d3a399dd299bf7e811cf42950d5f8ac67f063b36
> > Repository branch: native-comp
> > Windowing system distributor 'The X.Org Foundation', version 11.0.12010000
> > System Description: Debian GNU/Linux bullseye/sid
> >
> > Configured using:
> >  'configure --without-toolkit-scroll-bars --with-x-toolkit=lucid 
> > --with-nativecomp'
> >
> >
> > Debugger entered--Lisp error: (error "Lisp nesting exceeds 
> > ‘max-lisp-eval-depth’")
>
> Hi Michael,
>
> just tried compiling the current branch (f92bb788a0) following your
> configuration with no issues.  Could you retry on the current head maybe
> starting from a "git clean -xfd" to be on the safe side?

One thing I've noticed in my experiments is that many builds that are
interrupted at the wrong point and then resumed produce different
results. I.e. you type "make", then hit Ctrl-C at the wrong time, then
type "make" again and you get a different result.

We should fix that, and it might lead to the behavior Michael has
observed (macroexp being executed from source code when compiling
gnus)...

BTW, I'm also seeing very deep recursion when building the nativecomp
branch, even outside of the garbage collector (which is easy to "fix"
so it allocates more than one object at the same time on the C stack).
Maybe we can avoid that by doing an -O0 compile first, then compiling
the optimized compiler with the unoptimized one?