Closed fxcoudert closed 6 years ago
Ha, didn't realize they removed 2to3. While Python 2.7 is 2.5 years from final death and even the most entrenched are moving to Python 3, Apple is backing away.
(Feel free to delete this lament.)
There is now 2to3-
and 2to3-2.7
Oh yeah, I missed those. What the heck is 2to3-...
@zmwangx no idea. I could come up with theories but the only one that seems plausible is that someone hit the wrong key on their keyboard.
They do something funny with 2to3: https://opensource.apple.com/source/python/python-97.50.7/Makefile.auto.html
I guess 2to3-
isn't actually new though
No it's not. At least it's present on 10.12.
$ ls /usr/bin/2to3*
/usr/bin/2to3 /usr/bin/2to3- /usr/bin/2to3-2.7 /usr/bin/2to32.6
Going to mass replace 2to3 to 2to3- until I get around to converting those to patches https://github.com/Homebrew/homebrew-core/pull/18501 so that we can haz bottles.
Gearman doesn't seem to compile on high Sierra, as mentioned here: https://github.com/Homebrew/homebrew-php/pull/4447 so that can either be checked or crossed of.
I've tested bullet
on high sierra, and it builds from source fine. This issue says help wanted
, but I'm not sure how you build bottles for the new OS release. I haven't seen pull requests; I just see commits from @brewtestbot that are merged by a maintainer. Not sure if there's anything I can do.
@scpeters bullet
fails to build on our CI machines. This is due to b3GjkPairDetector.cpp
failing to compile with Xcode9's clang
when -target-cpu core2
(or -march=core2
) is passed.
@SMillerDev this list represents things that do not have bottles on High Sierra. So none of them built. If the box is checked that means there's no longer an issue, and the formula has a bottle now or the formula has been deleted. So gearman
being on the list without a check mark already conveys the fact that it doesn't build on 10.13.
Right, I thought it was just things that didn't happen to have one yet. My bad.
@SMillerDev ah, I see. No worries. Just didn't want you to think we were ignoring you :)
I found a hacky way to build libgda on High Sierra for anyone else who comes looking. It should also provide some insight into how to fix it properly:
1) Apply patch to /usr/local/opt/glib/bin/glib-mkenums
to deal with latin-1 encoded libgda files:
410a411,415
> if curfilename.startswith("gda-") or curfilename.startswith("libgda-"):
> encoding = "latin-1"
> else:
> encoding = "utf-8"
>
412c417
< curfile = io.open(curfilename, encoding="utf-8")
---
> curfile = io.open(curfilename, encoding=encoding)
2) In shell, make mysql libraries findable when libgda goes looking for them in the wrong places (replace 5.7.10
below with whichever mysql client version you have):
brew install mysql
cd /usr/local/include/
for f in ../Cellar/mysql-client/5.7.10/include/mysql/*.h; do ln -s "$f"; done
cd mysql
ln -s mysql/psi
for f in mysql/*.h; do ln -s "$f"; done
brew install libgda
@fxcoudert thanks for the info about bullet. It compiled fine with -march=native
, but I was able to reproduce the problem with -march=core2
. I tinkered with flags and it seems to be fine with -O1
and lower, but clang just takes forever with -O2
and higher. I've reported it upstream to bullet https://github.com/bulletphysics/bullet3/issues/1347
How are you building Bullet exactly? I'm using cmake or premake usually, and both work out-of-the-box fine with Bullet on High Sierra, but there isn't a flag -march=core2 in neither of those build systems. So how would I reproduce the issue using any of the default Bullet build systems?
Also, what is the output of clang --version?
@erwincoumans The bullet homebrew formula uses cmake and the -march=
compiler flags are supplied by brew to ensure support for a wide range of supported CPU architectures. We can continue discussion in https://github.com/bulletphysics/bullet3/issues/1347 unless you have specific questions for the homebrew maintainers.
fix for bdsup2sub https://github.com/Homebrew/homebrew-core/pull/18716 and a fix for passenger is coming but we're trying to squeeze in some other things to the release too.
@CamJN is there a passenger patch we can backport in the meantime?
@ilovezfs Not really, it's spread over a few commits and they aren't tested without the intermediary commits. The most important commit is this one: https://github.com/phusion/passenger/commit/416894542e3c689f4d609e6ce7f6c9a298ca3992 but I'm not sure if it can stand alone.
Note today's Passenger releases do not fix the High Sierra issue, they are due to a forced Enterprise release.
Apparently rake apache2
in the formula becomes /usr/local/Homebrew/Library/Homebrew/vendor/portable-ruby/2.3.3/bin/rake apache2
now? Even though /usr/local/Homebrew/Library/Homebrew/vendor/portable-ruby/2.3.3/bin/rake
doesn't exist? Can someone explain why you guys broke rake? We need rake to build Passenger, and I'm assuming you'll uphold your usual policy of not letting me fix this myself.
@CamJN please file an issue describing in detail your issue. On my 10.13 system it is correctly calling /System/Library/Frameworks/Ruby.framework/Versions/2.3/usr/bin/rake apache2
, which then fails:
clang -o buildout/apache2/mod_passenger.o -Isrc/agent -Isrc/cxx_supportlib -Isrc/cxx_supportlib/vendor-copy -Isrc/cxx_supportlib/vendor-modified -fPIC -D_REENTRANT -I/usr/local/include -Wall -Wextra -Wno-unused-parameter -Wno-parentheses -Wpointer-arith -Wwrite-strings -Wno-long-long -Wno-missing-field-initializers -fvisibility=hidden -DVISIBILITY_ATTRIBUTE_SUPPORTED -g -DHAS_ALLOCA_H -DHAS_SFENCE -DHAS_LFENCE -DPASSENGER_DEBUG -DBOOST_DISABLE_ASSERTS -c src/apache2_module/mod_passenger.c
src/apache2_module/mod_passenger.c:26:10: fatal error: 'httpd.h' file not found
#include <httpd.h>
^~~~~~~~~
1 error generated.
rake aborted!
I'm referring to a system that uses the new portable-ruby not the system ruby like one would expect. I know that passenger <= 5.1.10 won't compile on High Sierra but I was trying to test on el cap and found that portable-ruby is broken/crippled.
@CamJN if this is not on High Sierra, I am unclear why this is here. Please file an issue describing in detail your issue.
It's mentioned here because it has become one more yak to shave.
@CamJN Worry no more, that animal has been depilated. https://github.com/Homebrew/homebrew-core/pull/19376
It won't work, I said so in https://github.com/Homebrew/homebrew-core/issues/18493#issuecomment-336224014, But now I have to fight with the homebrew maintainers to fix their new bugs before I can work on getting the release that will fix High-Sierra out.
and this is how they behave: https://github.com/Homebrew/brew/issues/3310
@CamJN I am a Homebrew maintainer, and I did fix the rake
issue in that commit, less than an hour after you reported it above.
I guess now you can return to “work on getting the release that will fix High-Sierra out”, since High Sierra has been released for weeks (and months if we count betas)
Huh, so system "rake"
is allowed? That's the exact kind of thing that gets my PRs rejected all the time by you guys.
That's the exact kind of thing that gets my PRs rejected all the time by you guys.
??? Not in this universe
This is my final comment off-topic, and we'll try to focus this PR on getting High Sierra bottles!
@CamJN Please read, or re-read, our Code of Conduct and adjust your future communication accordingly.
I've read it, and I'm sorry I got angry. Homebrew is by far the most infuriating project I've ever interacted with, and because of the official and often enforced policy of not caring about the users I tend to read replies as having an air of condescension and dismissal that may not always be present.
I realize the policy isn't going to change, and that I have to live with it if I want to interact with the project at all, but being treated badly sucks; as you well know given you tell me not to do it to you all.
@CamJN Please stop being inflammatory, it is not constructive and helps no one.
The comment
because of the official and often enforced policy of not caring about the users
is entirely unnecessary, and furthermore is not factually accurate. It may be a personally held belief or opinion, but by including it in what is framed as a conciliatory statement perverts the apologetic tone into one that could be received in a number of quite negative ways.
Please, try to temper your tone and overall manner.
tinyproxy
seems to build fine now
For passenger
, just add dependence httpd
and apr
when building with apache2 module
Hi! I'm on High Sierra (version 10.13.1) and when I try to install OpenCV 3.4
by running the command brew install opencv --with-contrib
option, I got the warning from homebrew saying that Warning: opencv: this formula has no --with-contrib option so it will be ignored!
. I try doing brew options opencv
and nothing shows up. Is this because OpenCV 3.4.0
was only released several weeks ago and no updates have been made to homebrew?
Thanks!
it's always built with contrib now
Just created a PR that fixes ucspi-tcp
, meaning that there's just one more popular formula to go! (At least, if #22681 is accepted and merged. 😛)
What about htop
? (https://github.com/Homebrew/homebrew-core/issues/22622)
Might be related to https://github.com/hishamhm/htop/issues/682
Some updates:
Are we still doing depends_on MaximumMacOSRequirement
? Might be time to slap that on nethacked; the formula's already kind of a mess of hacks.
Are we still doing depends_on MaximumMacOSRequirement?
Only in cases where it's not feasible to backport a fix, but a fix is nonetheless expected or already merged, and at the same time an or_later
bottle won't work. For example: valgrind.
Patched ddd with PR #25233
Thanks to everyone who helped on this enormous task! We've got them all working… well, except a few we removed :)
A special thanks to @fxcoudert for his tireless and skillful work getting everything patched and bottled for High Sierra.
Remaining
ucspi-tcp
)Top formulas (> 200 installs a month)
(sorted by usage)
Done, 13 March 2018
clockid_t
)Other formulas
Done
librem
)globus-toolkit
)openmsx
)globus-toolkit
)solarus
)solarus
)solarus
)