cfbolz changed the topic of #pypy to: #pypy PyPy, the flexible snake | IRC logs: and | Matti: I made a bit of progress, the tests now only segfault towards the end
f4at has quit [Ping timeout: 244 seconds]
f4at has joined #pypy
f4at has quit [Client Quit]
<fijal> Heston: producing of pyc files is generally a side-effect that mostly does not have an effect on performance
_whitelogger has joined #pypy
krono has joined #pypy
krono has quit [Changing host]
krono has joined #pypy
* fijal is very confused
<fijal> do we support stacklet on OS X in general?
tazle has joined #pypy
Guest96_ has joined #pypy
shodan45 has quit [*.net *.split]
kurisu has quit [*.net *.split]
shodan45 has joined #pypy
<arigato> fijal: I find nothing that would disable continuations on OS X automatically
<fijal> arigato: yeah it seems that all was needed was to change the label from zero to 0
<fijal> ... because what got me as a surprise was that __GNUC__ is defined on clang
<arigato> ah
<arigato> of course it is
<fijal> yes, you can see the confusion :)
<Heston> fijal, alright. I just found it weird that there's a flag to disable them but they werent being produced by default
<Heston> or at least they don't seem to be
<fijal> yes, I would not pay much attention to them, why are they so interesting?
<fijal> mattip: branch might be done, I'm translating full pypy
<fijal> how do we set up the buildbot there?
<fijal> (with minor fixes to vmprof and continuation)
<mattip> I guess the question is which "there" are you referring to?
* LarstiQ points at the there
<fijal> yeah, that's part of the question
<fijal> do we have a possible m1 buildbot?
<LarstiQ> Heston: most of the interesting things happen _after_ the bytecode exists, there is little difference between cpython and pypy bytecode
<mattip> I think you have access to my machine still, I am fine with having it run a worker for M1
<mattip> and for x86_64 for that matter
<fijal> cfbolz: can we ask on twitter whether people are willing to sponsor buildbots for OS X?
<fijal> mattip: ok?
<mattip> sure, but the tweet should make it clear we want not only a machine but someone who will care and feed it as well
<fijal> yeah
<cfbolz> Didn't we write the info what we want down somewhere?
<mattip> the tricky part is getting python2.7 to run the worker
<fijal> mattip: pypy-c
<cfbolz> If somebody writes the tweet, I'm happy to post it from the PyPy account, or retweet
<cfbolz> Also, maybe a PyPy dev mail?
<fijal> cfbolz: "We are looking for people willing to run and maintain a pypy buildslave for OS X (both M1 and x86)"
<fijal> or am I not poetic enough?
<fijal> mattip: ok, I think I'm done?
<fijal> until we have a buildbot at least
<cfbolz> I can ask whether we have a machine at the uni we could use, but that would need administering
<fijal> I'm not volunteering
<fijal> ok, merged m1
<arigato> yay!
<fijal> cfbolz: can you tweet?
<cfbolz> fijal: err, did you run any tests on the m1 branch on linux and maybe mac-x86 to see that nothing else breaks?
<fijal> I did run some, not all
<fijal> but not on buildbot, let me do that
<fijal> well, I can't run them on mac
<cfbolz> fijal: what's the perf impact of all the enter/leave_assembler_writing on say, linux on a raspi?
<fijal> cfbolz: nothing? those are empty functions
<cfbolz> ok, still in the middle of reading the diff
<fijal> I run some benchmarks - seems like they're very fast
<fijal> (just sets some flags)
<cfbolz> fijal: what about the ClosureHeap freelist removal?
<cfbolz> I suppose on mac you can't use that
<cfbolz> but what on other OS?
<fijal> I explained earlier - seems like using free list on libffi makes no sense
<fijal> they do allocate their own freelist on systems where it makes sense
<cfbolz> ok, good
<cfbolz> no, you didn't explain :-P
<fijal> that was my read from the massive closures.c file
<cfbolz> you just said "look on mac os m1 there is no freelist" ;-)
<fijal> heh
<fijal> mattip: do you know how do I submit the invoice? who do I invoice?
Guest96_ has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
slav0nic has joined #pypy
otisolsen70 has joined #pypy
<mattip> nope, I don't know if we have ever paid out of OpenCollective
<mattip> maybe ronan or antocuni ?
Guest96_ has joined #pypy
<antocuni> I have never been paid by OpenCollective
<cfbolz> he figured it out
<fijal> cfbolz: I think I did not break extra tests?
<cfbolz> great!
<cfbolz> let's see tomorrow too
<cfbolz> fijal: can you also merge to py3.8? should apply cleanly, but you never know
<fijal> how? like hg merge default?
<cfbolz> yeah
<fijal> scary
Guest96_ has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<fijal> cfbolz: like really hg merge default on py3.8? I would not be willing to review the diff
<cfbolz> fijal: the diff is like the diff of your merge to default, you touched very few things. however, there is this:
<cfbolz> you touched few things in pypy/ I mean
<cfbolz> however (I just tried): there is one py3.8 change in rpython/ that's relevatn
<cfbolz> rpython/translator/platform/
<cfbolz> has: DARWIN_VERSION_MIN = '-mmacosx-version-min=10.7'
f4at has joined #pypy
<fijal> right, I think we should bump it?
<cfbolz> I have no opinion on that ;-)
<fijal> 10.7 is ancient and OS X users are quite good at updating
<fijal> (like 11 years old)
<cfbolz> but it seems if you bump it, you also need to bump it in lib_pypy/
<cfbolz> as per the comment
<cfbolz> 12.0 is really quite recent. but as I said, I have no opinion and no clue
<fijal> 10.13?
reneeontheweb has joined #pypy
f4at has quit [Ping timeout: 260 seconds]
f4at has joined #pypy
f4at has quit [Remote host closed the connection]
<mattip> fijal: could you update the documentation on how to build for M1, in particular
<mattip> I cannot build since it cannot find expat, even though I brew installed it and set PKG_CONFIG_PATH
<fijal> I don't know what to do with that? it just worked for me....
<mattip> I don't know what to do either
<fijal> my point being that I can't update information on stuff I don't know about :)
<fijal> mattip: removing the x86 homebrew helped a lot btw
<mattip> ok, so that is one hint that might be nice to mention in the documentation
<fijal> ok, I can update that for sure
<mattip> is the platform name supposed to be "Darwin_arm64"?
<fijal> arm64 or aarch64
reneeontheweb has quit [Quit: Client closed]
<fijal> the x86 simulation is REALLY IMPRESSIVE
<fijal> like, gcbench runs at the same speed
<fijal> richards is 30% faster
<mattip> host_factory = "Darwin_arm64" so that is not the problem
<fijal> mattip: I don't think Im following
<mattip> I am trying to figure out how pyexpat is supposed to find the library. It uses platform.preprocess_library_dirs but what is platform?
<mattip> for macmini it is Darwin_arm64
<mattip> and the directory is empty
<fijal> I have had a very hard time running anything on your macmini, I don't know why :/
<mattip> could you add a pdb there and tell me what it is in your successful translation?
<mattip> interp_pyexpat
<fijal> (I don't have the same problems on my machine)
<fijal> mattip: sorry?
<LarstiQ> fijal: your machine is a laptop?
<mattip> we only have two data points: your laptop that works and mine that doesnt. Can you help figure out what is wrong?
<fijal> LarstiQ: yes?
<fijal> mattip: I don't know :) I could not figure out why your macmini does not work
<fijal> (like I failed and gave up)
<mattip> in your machine, inside there is a line that finds the libraries and include files. How does your machine do it?
<fijal> so you want to see where my interp_pyexpat takes the libraries from?
<mattip> yes
<mattip> so that I can compare to mine, which are empty
<mattip> even better would be a set of instructions so other people can duplicate your work
<fijal> mattip: I don't think there are special instructions
<fijal> I do believe you did something magic on your machine which did something that made it not work
jacob22 has quit [Quit: Konversation terminated!]
<mattip> I believe you did something magic on your machine to did something that made it work
<fijal> well, if I did, then I have no idea what
<fijal> I installed homebrew
<mattip> how exactly?
<fijal> by googling "how to install homebrew" and doing that
<fijal> and then setting the PATH and stuff
<mattip> cool. I will wipe my machine and try again. Could you share your PATH and stuff?
<fijal> again, it was just following the homebrew
<fijal> x /Library/Developer/CommandLineTools/SDKs/MacOSX11.3.sdk/System/Library/Frameworks/Python.framework/Versions/2.7/include/python2.7/pyexpat.h
<fijal> sorry
<fijal> x /Library/Developer/CommandLineTools/SDKs/MacOSX12.3.sdk/usr/include/expat.h
<fijal> this is where my expat.h lives, which seems like a standard place?
<fijal> what does your gcc -v give?
<mattip> nothing right now, I am wiping the machine and starting over
<fijal> ok
<fijal> I can try to help, but I really don't know how you got your machine in that state
jacob22 has joined #pypy
<fijal> looks like expat.h should come from the command line tools SDK
slav0nic has quit [Ping timeout: 258 seconds]
<mattip> thanks, I will let you know once I have reset everything and installed homebrew like you did
<fijal> mattip: cool, I updated the docs, I think?
<mattip> this version on HEAD still says to use x86_64 emulation
lritter has joined #pypy
slav0nic has joined #pypy
slav0nic has quit [Ping timeout: 255 seconds]
jacob22 has quit [Ping timeout: 258 seconds]
slav0nic has joined #pypy
Guest96 has joined #pypy
reneeontheweb has joined #pypy
Guest96 has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
slav0nic has quit [Ping timeout: 240 seconds]
f4at has joined #pypy
slav0nic has joined #pypy
Guest96 has joined #pypy
otisolsen70 has quit [Quit: Leaving]
Guest96 has quit [Quit: Textual IRC Client:]
Atque has joined #pypy
jacob22 has joined #pypy
f4at has quit [Ping timeout: 256 seconds]
<fijal> uh wat?
Atque_ has joined #pypy
Atque has quit [Ping timeout: 240 seconds]
Atque_ is now known as Atque
Atque has quit [Remote host closed the connection]
Atque has joined #pypy
lritter has quit [Quit: Leaving]
<mattip> gcc -v gives me lots of output, including
<mattip> Apple clang version 13.0.0 (clang-1300.0.27.3)
<mattip> fijal: how did you get a cpython2.7 for bootstrapping translation? pyenv?
<cfbolz> graaaah, hash(int) is producing lots of guards and bridges in pypy3?!
slav0nic has quit [Ping timeout: 260 seconds]