cfbolz changed the topic of #pypy to: PyPy, the flexible snake (IRC logs: https://quodlibet.duckdns.org/irc/pypy/latest.log.html#irc-end ) | use cffi for calling C | if a pep adds a mere 25-30 [C-API] functions or so, it's a drop in the ocean (cough) - Armin
dddddd has quit [Read error: Connection reset by peer]
ajlawrence has quit [Remote host closed the connection]
ajlawrence has joined #pypy
<ajlawrence>
arigato: what is confusing in this instance is that I do not call charp2str)
<arigato>
the message says "some variable can store either a W_UnicodeObject or an rpython unicode"
<arigato>
indeed it's strange
<arigato>
ah, found it
<arigato>
you have "return rffi.wcharp2unicode(buf)" => rpython unicode
<arigato>
but also "return space.newutf" => W_UnicodeObject
<ajlawrence>
Ah yes thanks, I was looking at the definition for wcharp2unicode from the rawffi module which has return space.newutf8(...) I think that helped confuse me.
<arigato>
you need two lines instead, just like in this rawffi function you saw
<arigato>
the first one should call unicodehelper.wcharpsize2utf8, instead of directly rffi.*
ajlawrence has quit [Remote host closed the connection]
extraymond[m] has joined #pypy
lritter has joined #pypy
antocuni has joined #pypy
forgottenone has quit [Remote host closed the connection]
forgottenone has joined #pypy
altendky has quit [Quit: Connection closed for inactivity]
<kenaan>
arigo cffi/cffi 3873bbdd3b9e /: Issue #429 There are corner cases in which we can see a recursion on the same types. Instead of fighting them al...
<kenaan>
arigo default 847f17dd11a6 /pypy/module/_cffi_backend/: manual update corresponding to cffi/3873bbdd3b9e
<kenaan>
arigo default ec4f85880b7c /: update all the way to cffi/3873bbdd3b9e
<kenaan>
arigo cffi/cffi dd09b2809748 /: bump version to 1.13.1
<kenaan>
arigo default b514063ad22a /: bump cffi version to 1.13.1
forgottenone has quit [Ping timeout: 240 seconds]
forgottenone has joined #pypy
lastmikoi has joined #pypy
jcea has joined #pypy
YannickJadoul has joined #pypy
jcea has quit [Remote host closed the connection]
antocuni has joined #pypy
Dejan has quit [Quit: Leaving]
jcea has joined #pypy
jvesely has joined #pypy
Rhy0lite has joined #pypy
Dejan has joined #pypy
YannickJadoul has quit [Ping timeout: 252 seconds]
<arigato>
Alex_Gaynor: can you trigger a build for CPython 3.8? right now the azure pipelines thing is great but doesn't contain 3.8 yet
<arigato>
if you need me to release in source form first, then I'll release the source and the wheels that I have so far, and only add the 3.8 wheels from pypa
<Alex_Gaynor>
arigato: ah, yes, it builds whatever the latest release from PyPI is, so if you haven't uploaded it yet, it won't work
<arigato>
OK, will upload and ask you again maybe in 30min
<arigato>
lunch
<Alex_Gaynor>
ok
<antocuni>
what are the differences/advantages of azure pipelines vs travis? Apart the obvious fact that travis works only with github
<Alex_Gaynor>
antocuni: well, until a few weeks ago, azure had macOS/linux/Windows, but now travis has all of them too
<antocuni>
ah, but I see it works only with git? So no chances to use it for pypy
marky1991 has quit [Remote host closed the connection]
<arigato>
I'm unsure what you have in mind, but if it's to run the complete tests every time there's a commit, I think we'd blow up all the way in the very expensive plans
jcea has quit [Remote host closed the connection]
<Alex_Gaynor>
arigato: FWIW, MS has been very nice about giving us free services. I don't know exactly what the limits would be, but they gave PyCA 50 concurrent builders just because we asked
<antocuni>
I'm not sure. On our buildbot, own-linux-64 tests take ~1h40min
<antocuni>
assume it's 40 minutes for rpython and 1hr for "real" pypy tests, (assuming that we finally split the two repos)
<antocuni>
and this is another random project which runs tests on an endless combination of python/pyest/whatever versions; total time "more than 24 hrs": https://travis-ci.org/ionelmc/pytest-benchmark
<antocuni>
(real clock time: 6 hours)
<antocuni>
and it happens on every push, not every commits
<Alex_Gaynor>
antocuni: I know MS sponsors rust with buidlers that hvae a 4 hour timelimit and like 50 concurrency, so it's possible, just need to find a friendly person to ask. But it's still git only, so maybe not helpful
<antocuni>
arigato: also, as Alex_Gaynor says, I think that nowadays the limiting factor for using build services is the number of parallel jobs, not the total time which it takes. But even with the default free plan for open source projects, azure gives you 10 parallel jobs, which means ~10 minutes wall-clock-time to run the whole pypy suite, if we configure the jobs optimally
<antocuni>
I wonder if it would work to have a git mirror whose job is only to trigger CI builds :)
<arigato>
mattip: if you ask me, trying to run some subset of tests (like cpyext) on Azure looks much more convincingly like the real thing
<arigato>
(or Travis etc.)
<mattip>
how would you advise doing that? Can I put `hg clone` into an azure.yml command?
<arigato>
right, it's probably messy
<arigato>
based on my extremely limited experience, I would make a dummy git repo with the azure.yml and commit something random to it every time I want to run the tests; and yes you should be able to put 'hg clone'
<arigato>
maybe with "python -m pip install hg"
<mattip>
ok, worth thinking about
<arigato>
you don't get the same thing as with an emulator running on your machine
<arigato>
as in, it takes a while to run and you have no hope of debugging things
<mattip>
well, it is not clear I will get that via the darling route anyway :)
antocuni has quit [Ping timeout: 264 seconds]
BPL has joined #pypy
YannickJadoul has joined #pypy
<mattip>
fwiw, I am trying to track down the timeout on macos py3.6 -A testing, search for "! module/cpyext/test" on