<ajlawrence>
The windows py3.6 builds are failing. Does anyone have an idea what the root cause is there? There is a warning about _hashlib not being available and that there is no module named _pypy_openssl.
<mattip>
I think it has to do with not being able to build blake2. We started to fix it for linux32 in 323ddcb8a8c4 but maybe it still fails on win32
<mattip>
with some kind of crash when raising an error in _winreg. Maybe it will show up in untranslated tests?
<mattip>
nice. Untranslated tests segfault the host python2
<mattip>
down the rabbit hole we go
<ajlawrence>
When I try _blake2_build.py I get blake2-config.h(70): fatal error C1189: #error: " This code requires at least SSE2." followed by a number of exceptions.
<ajlawrence>
Is the problem there or is it deeper within the code base?
<mattip>
you need a newer pypy3, that was before 323ddcb8a8c4
<mattip>
but now I see there is a problem with some rpython code in interp_winreg.py, so that needs fixing too
<mattip>
so I guess the answer is "deeper in the code base"
<kenaan>
mattip py3.6 28ca4c3a2982 /lib_pypy/_blake2/_blake2_build.py: fix 323ddcb8a8c4 for win32 which needs to define the macro, sse2 is always supported
demonimin has quit [Remote host closed the connection]
<amitripshtos>
Hey guys, I'm trying to get into pypy codebase in order to help maintain the project. Any tips on getting started? Maybe any known bugs that are easy to reproduce and solve? Thanks
<simpson>
amitripshtos: It's often good to start with something that matters to the contributor. What are some problems that *you* have experienced with PyPy?
<amitripshtos>
<simpson>: Actually I really want to see pypy works with python3.6 (and 3.7 for the dataclasses and named enum typing), not sure how hard it is to contribute to it
<simpson>
amitripshtos: Totally understandable. New Python 3.x features are a common thing for folks to want to work on in PyPy.
<simpson>
It seems, from mattip's link above, that the `py3.6` branch is the one to work on and pay attention to.
<mattip>
if there are pure-python implementations, it might be as simple as updating the stdlib and making sure tests pass
<mattip>
for dataclasses and named enum
<amitripshtos>
Great, thanks! I will start looking at the issues.
<mattip>
cool. It is easy to get lost in the details, feel free to ask questions before diving in too deep
k1nd0f has joined #pypy
k1nd0f has quit [Client Quit]
<ajlawrence>
mattip: It looks like the windows failed to complet. Atleast it did not upload the executables.
<ajlawrence>
I ran the translation locally on my machine and tried the test_asyncio\test_windows_events.py tests and now get a crash.
<ajlawrence>
..RPython traceback: File "pypy_interpreter.c", line 39090, in BuiltinCode_funcrun_obj File "implement_6.c", line 39957, in BuiltinActivation_UwS_W_FFIObject_int__run Fatal RPython error: AssertionError
mattip has quit [Ping timeout: 255 seconds]
kostia has joined #pypy
<arigato>
ajlawrence: I can help you debug that, if you want
<ajlawrence>
arigato: I have no idea where to start there so any help is greatly appreciated.
<arigato>
the error means "there is some assert in the source code that triggered"
<arigato>
to learn more, you should try to reproduce the assert on an untranslated pypy
<ajlawrence>
How do I tell which assert?
<arigato>
it's somewhere inside W_FFIObject.int(), I think
<ajlawrence>
What is the best way to reproduce the assert? are there some tests I can run?
<arigato>
yes, but I don't know if these will find the failure
<arigato>
to reproduce, first try to isolate a program sample that fails in the same way immediately
<arigato>
then run this same program on top of an untranslated pypy
<arigato>
where you can use pdb and all other tools you'd normally use to debug Python (if needed)
<Ninpo>
I had a problem, so I decided to use regex, now I have \d+ problems
<arigato>
:-)
<arigato>
(note: "run untranslated" here can be done with: python2 pypy/bin/pyinteractive.py -S file.py)
<arigato>
(or pypy2 instead of python2)
<Ninpo>
Wrong channel, sorry to spam with a bad joke
<arigato>
that's not a problem here, don't worry
amitripshtos has quit [Quit: Leaving]
<ajlawrence>
arigato: So with pyinteractive.py I can run bits of untranslated code and see what happens?
<arigato>
yes
mattip has joined #pypy
<arigato>
it's extremely slow, so use only if you isolated a script that shows the problem immediately without importing large libraries
<mattip>
ajlawrence: did that test not crash before?
<ajlawrence>
Nope
<mattip>
sorry, too many negatives. Did that test fail/succeed without a crash :) ?
<ajlawrence>
There were a mixture of failures and successes previously on my branch. I believe they all failed previously in the py3.6 branch.
<mattip>
about the failed upload, it seems we need to build liblmza in-place, and not prebuild it, there are differences between msvc 14.0 versions (maye)
<mattip>
maybe
<ajlawrence>
I had a similar issue on my machine and it seemed to be getting confused between VS 2015 and 2017 when retrieving the externals.
kostia has quit [Remote host closed the connection]
<ajlawrence>
I tried running pypy.exe pytest.py -A -rs --python=python pypy\module\_cffi_backend\test\test_ffi_obj.py and got the error ModuleNotFoundError: No module named '_cffi_backend'
<mattip>
add --withmod-_cffi_backend, and then you need to add lots of other modules as well to make that work
<mattip>
--help
<mattip>
ahh, no erase that
<mattip>
python2 pytest.py -A -rs --python=path/to/pypy3.exe pypy\module\...
<mattip>
running with -A is running tests after translation, so like running plain python pytest testfile, but
<mattip>
our testing is all python2, so we need to play games: we write out a file that contains the test, and run that directly
<mattip>
untranslated tests are without the -A, so "python2 pytest.py pypy\module\_cffi_backend\test\test_ffi_obj.py"
<mattip>
which translates the modules needed to run the test and then runs it untranslated, so you can put in pdb and stop in rpython code
<ajlawrence>
I have run that command. Lets see what happens...
<mattip>
I think arigato was suggesting a third mode: to run the interpreter untranslated, and let it try a small file that crashes, that mode is
<mattip>
his command from above
<ajlawrence>
My problem is that I do not have much of an idea what is causing the crash currently so I am not sure the third mode is possible for me.
<ajlawrence>
pypy\objspace\std\unicodeobject.py:47: AssertionError ========================== short test summary info =========================== FAIL pypy/module/_cffi_backend/test/test_ffi_obj.py::AppTestFFIObj::()::test_cha r32_t ==================== 1 failed, 45 passed in 77.59 seconds ====================
<ajlawrence>
One failure there.
<mattip>
can you paste the AssertionError in a pastebin (too many lines for here)?
<mattip>
there should be a call to newtext somewhere in the stacktrace
<mattip>
on python3 linux64, but 2 on python2 win32
<mattip>
not good
<mattip>
ajlawrence: that code is only executed untranslated, your crash was in translated code. So while that error is not good, I do not think it is causing the crash
nunatak has quit [Remote host closed the connection]
jacob22 has quit [Read error: Connection reset by peer]
jacob22 has joined #pypy
mattip_ has joined #pypy
mattip has quit [Ping timeout: 245 seconds]
<kenaan>
stevie_92 cpyext-gc-cycle 3ab559b9b893 /: First version of rawrefcount legacy finalizer implementation Test is still failing, because garbage_pyp...
ajlawrence has quit [Ping timeout: 256 seconds]
lritter has quit [Quit: Leaving]
Ai9zO5AP has quit [Quit: WeeChat 2.3]
k1nd0f has joined #pypy
xcm has quit [Killed (hitchcock.freenode.net (Nickname regained by services))]