FS#74433 - [emacs-nativecomp] Needs 'make bootstrap' to make NATIVE_FULL_AOT=1 effective
Attached to Project:
Arch Linux
Opened by Thomas Eberhardt (ed209) - Sunday, 10 April 2022, 11:56 GMT
Last edited by freswa (frederik) - Wednesday, 20 April 2022, 17:11 GMT
Opened by Thomas Eberhardt (ed209) - Sunday, 10 April 2022, 11:56 GMT
Last edited by freswa (frederik) - Wednesday, 20 April 2022, 17:11 GMT
|
Details
Since the emacs .tar.xz distribution file contains all the
byte-compiled .elc files
"make NATIVE_FULL_AOT=1" doesn't have the intended effect of pre-compiling all .el files natively. You need to do a full bootstrap, and since zlib is always available (at a minimum, cairo sucks it in) "--without-compress-install" is also not needed. See attached PKGBUILD diff. |
This task depends upon
Closed by freswa (frederik)
Wednesday, 20 April 2022, 17:11 GMT
Reason for closing: Implemented
Additional comments about closing: emacs-28.1-4 in [testing]
Wednesday, 20 April 2022, 17:11 GMT
Reason for closing: Implemented
Additional comments about closing: emacs-28.1-4 in [testing]
Though, why should we remove `--without-compress-install`? I do agree with the arguments of the last thread, that compression should be done on the filesystem level. These files are usually frequently read and performance depends on the system. So setting a global standard seems inconvinient to me.
When you do build with "make NATIVE_FULL_AOT=1 bootstrap" there is an .eln & .elc file built for every
.el file, so I think there aren't ever going to be accessed by emacs, and are there just for reference,
which is probably the reason there are by default installed as compressed files.
I don't what to open a can of worms :-), but if one goes down the road that compression should be done
by the filesystem, shouldn't than all man-pages and info files and other compressed stuff under
/usr/share be installed uncompressed by all packages?
Arggh... what have I done :-) Just forget the stuff about "--without-compress-install"...
Even by people who RTFM, these files are accessed once to be loaded into memory. So the access times don't really matter opposed to code that needs to be loaded at runtime.
> so I think there aren't ever going to be accessed by emacs, and are there just for reference
I have to investigate that. A link to documentation regarding these files would be helpful.
If it's true, compression is a no-brainer for sure :)
$ info elisp "how programs do loading"
.elc < .eln > .el