--- Begin Message ---
Subject: |
28.0.50; [PATCH] 28.0.50; Native compilation unnecessarily recompiles .eln (macOS) |
Date: |
Sat, 12 Jun 2021 22:13:13 -0500 |
This is similar to bug#48497, but appears to still happen even after
commit 3f207753a0.
The basic problem is that the installed lisp path does not match either
of the search expressions in comp-el-to-eln-rel-filename, meaning that
native lisp needs to be recompiled needlessly.
The problem seems to come from PATH_REL_LOADSEARCH being set for me (on
macOS) to Contents/Resources/lisp, but the lisp files actually being
installed to
/nix/store/...-emacs-gcc-20210612.0/share/emacs/28.0.50/lisp (path
generated by Nix). As a result, comp-el-to-eln-rel-filename used the
filename comp-034d3699-516ce4bf.eln for comp.el.gz where 516ce4bf is the
md5sum of the contents of comp.el and 034d3699 is the md5sum of the full
path of comp.el, not of //emacs-lisp/comp.el (7672a6ed), which is what
Emacs installs.
To fix this, I propose using PATH_LOADSEARCH in addition to
PATH_REL_LOADSEARCH so that we can catch both types of macOS installs
(.app and unix). I’ve attached a patch which implements this, adding
relative and absolute loadsearch resolution.
- Matthew
>From 32022ee8d977196c72ad42d89d23142a6e59ff8e Mon Sep 17 00:00:00 2001
From: Matthew Bauer <mjbauer95@gmail.com>
Date: Sat, 12 Jun 2021 00:37:28 -0500
Subject: [PATCH] Use PATH_LOADSEARCH to get absolute path of native comp
We need to so that
/usr/local/share/emacs/28.0.50/lisp/emacs-lisp/comp.el
becomes:
//emacs-lisp/comp.el
which is what we install eln files as.
---
src/comp.c | 10 ++++++----
1 file changed, 6 insertions(+), 4 deletions(-)
diff --git a/src/comp.c b/src/comp.c
index 056d0860d8..8ccd99d583 100644
--- a/src/comp.c
+++ b/src/comp.c
@@ -4049,12 +4049,14 @@ DEFUN ("comp-el-to-eln-rel-filename",
Fcomp_el_to_eln_rel_filename,
As installing .eln files compiled during the build changes their
absolute path we need an hashing mechanism that is not sensitive
- to that. For this we replace if match PATH_DUMPLOADSEARCH or
- *PATH_REL_LOADSEARCH with '//' before computing the hash. */
+ to that. For this we replace if match PATH_DUMPLOADSEARCH,
+ PATH_REL_LOADSEARCH, or PATH_LOADSEARCH with '//' before
+ computing the hash. */
if (NILP (loadsearch_re_list))
{
- Lisp_Object sys_re =
+ Lisp_Object sys_abs_re = Fregexp_quote (build_string (PATH_LOADSEARCH
"/"));
+ Lisp_Object sys_rel_re =
concat2 (build_string ("\\`[[:ascii:]]+"),
Fregexp_quote (build_string ("/" PATH_REL_LOADSEARCH "/")));
Lisp_Object dump_load_search =
@@ -4062,7 +4064,7 @@ DEFUN ("comp-el-to-eln-rel-filename",
Fcomp_el_to_eln_rel_filename,
#ifdef WINDOWSNT
dump_load_search = Fw32_long_file_name (dump_load_search);
#endif
- loadsearch_re_list = list2 (sys_re, Fregexp_quote (dump_load_search));
+ loadsearch_re_list = list3 (sys_abs_re, sys_rel_re, Fregexp_quote
(dump_load_search));
}
Lisp_Object lds_re_tail = loadsearch_re_list;
--
2.29.2
--- End Message ---
--- Begin Message ---
Subject: |
Re: bug#48994: 28.0.50; [PATCH] 28.0.50; Native compilation unnecessarily recompiles .eln (macOS) |
Date: |
Sat, 26 Jun 2021 10:38:19 +0100 |
I've pushed it to master.
Thanks!
On Wed, Jun 23, 2021 at 03:46:12PM -0500, Matthew Bauer wrote:
> That works for me! No apparent issues.
>
> On Wed, Jun 23, 2021 at 10:59 AM Alan Third <alan@idiocy.org> wrote:
>
> > On Tue, Jun 22, 2021 at 06:01:25PM -0500, Matthew Bauer wrote:
> > > > Turns out there's another variable that should only be set when
> > > > building the app.
> > >
> > > > Please try the attached patch.
> > >
> > > That works for me !
> > >
> > > So previously the nextstep/Emacs.app was getting generated even
> > > with --disable-ns-self-contained. I think that's fine to not build in
> > > this case - in fact it duplicates the Emacs executable - but just a note
> > > that it kind of changes things for packages. I have a fix for Nixpkgs
> > > (which previously installed nexstep/Emacs.app), but I think
> > > homebrew-emacs-plus would also be effected:
> > >
> > > https://github.com/d12frosted/homebrew-emacs-plus
> >
> > This sounds like madness to me, but I see it's a supported
> > configuration mentioned in our install files.
> >
> > So, attempt three attached!
> >
--
Alan Third
--- End Message ---