Compiler optimization flag: -O2

[stradicat]

Added -O2 to CFLAGS for some optimizations on the C side.
I've tested in macOS, and used it for a long time in Linux and BSD, both on amd64 and i386 architectures, to improve responsiveness.

[stradicat]

I get a test error specifically with emacs-plus@29 and --with-native-comp enabled:

Error: Failed changing dylib ID of /usr/local/Cellar/emacs-plus@29/29.4/Emacs.app/Contents/native-lisp/29.4-8d609dc9/preloaded/oclosure-b279097d-98bf8ff8.eln

All other versions --with-native-comp passed the tests.

Could it be a pipeline issue, or something else?

[ilyaw39]

I get a test error specifically with emacs-plus@29 and --with-native-comp enabled:

Error: Failed changing dylib ID of /usr/local/Cellar/emacs-plus@29/29.4/Emacs.app/Contents/native-lisp/29.4-8d609dc9/preloaded/oclosure-b279097d-98bf8ff8.eln

All other versions --with-native-comp passed the tests.

Could it be a pipeline issue, or something else?

I also got this strange error while testing on CI. But this seems does not affect installations locally though as seen in #691. Possible solutions may include using gln in coreutils and using newer versions of clang SDK that llvm provides. This CI-related error can be ignored anyway.

---

Besides that, may I take a look at your M-x emacs-build-description? Mine is

...

--with-ns --disable-ns-self-contained 'CFLAGS=-Os -w -pipe
 -mmacosx-version-min=15 -isysroot/Library/Developer/CommandLineTools/SDKs/MacOSX15.sdk
 -O2 -march=native -DFD_SETSIZE=10000 -DDARWIN_UNLIMITED_SELECT

...

After a similar modification on the .rb file like yours. Note -Os is added to CFLAGS by default. This probably is done by Makefile that Emacs provides and overrides the -O2 option.

[stradicat]

I get a test error specifically with emacs-plus@29 and --with-native-comp enabled:
Error: Failed changing dylib ID of /usr/local/Cellar/emacs-plus@29/29.4/Emacs.app/Contents/native-lisp/29.4-8d609dc9/preloaded/oclosure-b279097d-98bf8ff8.eln
All other versions --with-native-comp passed the tests.
Could it be a pipeline issue, or something else?

I also got this strange error while testing on CI. But this seems does not affect installations locally though as seen in #691. Possible solutions may include using gln in coreutils and using newer versions of clang SDK that llvm provides. This CI-related error can be ignored anyway.

Besides that, may I take a look at your M-x emacs-build-description? Mine is

...

--with-ns --disable-ns-self-contained 'CFLAGS=-Os -w -pipe
 -mmacosx-version-min=15 -isysroot/Library/Developer/CommandLineTools/SDKs/MacOSX15.sdk
 -O2 -march=native -DFD_SETSIZE=10000 -DDARWIN_UNLIMITED_SELECT

...

After a similar modification on the .rb file like yours. Note -Os is added to CFLAGS by default. This probably is done by Makefile that Emacs provides and overrides the -O2 option.

The modified formula in my PR builds locally without problems, so the issue is on the GH pipeline side.

My Emacs M-x emacs-build-description shows:
--with-ns --disable-ns-self-contained 'CFLAGS=-Os -w -pipe -march=westmere -mmacosx-version-min=13 -isysroot/Library/Developer/CommandLineTools/SDKs/MacOSX13.sdk -O2 -DFD_SETSIZE=10000 -DDARWIN_UNLIMITED_SELECT'

(I'm on Ventura)

-O2 is the only CFLAGS modifier that doesn't crash the build, as opposed to the likes of -ftree-parallelize-loops=n and so on.

[d12frosted]

looks nice, thanks!
and yeah, ignore the with-native-comp on Emacs 29