#pypy on 2017-05-28 — irc logs at freenode.irclog.whitequark.org

14:48 <kenaan> arigo default 844e78f165f3 /lib-python/2.7/ctypes/test/test_unaligned_structures.py: Skip the test about byte-swapped structures, for now

14:50 <mattip_> it seems PyPy_profopt_enabled touched app_main

14:50 <mattip_> somehow causing http://buildbot.pypy.org/summary/longrepr?testname=AppTestAppMain.%28%29.test_setup_bootstrap_path&builder=own-linux-x86-32&build=5211&mod=pypy.interpreter.test.test_app_main

14:51 <mattip_> and I have no idea whether test_posix_fadvise, test_posix_fallocate failures are true failures or the test is wrong

14:52 <arigato> looks like a hack, to look harder in case we don't find lib-python and lib_pypy

14:52 <arigato> might cause more confusion than help, imho

14:56 <arigato> we should try to get cfbolz or mihai to explain it again

14:56 mattip_ is now known as mattip

14:57 <cfbolz> arigato: the problem is that pypy-c didn't run from the usession dir any more

14:57 <cfbolz> Imo remove this part and replace it with whatever other hack that makes that work

14:58 <arigato> the problem is only that sys.prefix is not set, right?

14:58 <cfbolz> I don't know

14:58 <arigato> well, if you don't know, then I have nothing to do :-) pypy does run from the usession dir

14:59 <arigato> but some specific things don't run

14:59 <arigato> so I need to know which ones

14:59 <arigato> as far as I remember, there is "import sysconfig" which fails because sys.prefix is not set

15:00 <cfbolz> arigato: try running the CPython tests from there

15:00 <arigato> ok

15:05 <arigato> yes, seems to mostly work

15:05 <arigato> (a few tests are confused by the output to stderr)

15:05 <cfbolz> arigato: more precisely, those tests: test_asyncore test_gdb test_multiprocessing test_subprocess

15:06 <arigato> test_gdb cannot work

15:06 <arigato> SkipTest: test_gdb only works on source builds at the moment.

15:07 <cfbolz> h

15:07 <cfbolz> ugh

15:07 <cfbolz> indeed, it's a bit nonsensical to use that as training

15:07 <arigato> uh?

15:07 <arigato> if you ask me, test_multiprocessing and test_subprocess are also nonsensical

15:07 <cfbolz> yes

15:08 <cfbolz> well, unclear: I think what pgo "only" does is speed up machine code generation

15:08 <cfbolz> so tests are the right training

15:08 <arigato> for the JIT?

15:08 <cfbolz> yes, because the rpython code of the jit gets better optimized

15:09 <cfbolz> (that's my theory at least)

15:09 <arigato> are these tests really calling the JIT?

15:09 <cfbolz> arigato: even importing site.py runs the jit

15:09 <arigato> yes, but my question was "are they doing that a lot"

15:10 <cfbolz> asyncore probably is

15:10 <cfbolz> but yes, the others look a bit silly

15:15 <arigato> I'd suggest something like "pypy-c pytest.py pypy/objspace/test/"

15:15 <cfbolz> sounds good to me

15:16 <arigato> it runs both a number of short tests and a couple of long-running ones

15:20 <cfbolz> arigato: one question is whether we want to do release build with profopt

15:22 <arigato> good question

15:22 <arigato> so it really speeds up the JIT?

15:23 <cfbolz> it seems so

15:25 aboudreault has quit [Excess Flood]

15:29 aboudreault has joined #pypy

15:33 aboudreault has quit [Excess Flood]

15:33 <arigato> bah, translation fails right now

15:33 <cfbolz> :-(

15:33 <cfbolz> because?

15:33 <arigato> http://paste.pound-python.org/show/LyupsNXSmx0SE27tk5cQ/

15:34 <cfbolz> "ah" :-(

15:34 <arigato> ah, it's 0e27a73c87f6

15:35 <ronan> oops

15:35 <arigato> np

15:35 <arigato> buf.array => buf.w_array, right?

15:35 <ronan> yes

15:37 aboudreault has joined #pypy

15:41 aboudreault has quit [Excess Flood]

15:42 aboudreault has joined #pypy

15:45 <kenaan> arigo default 1e656d2369ef /pypy/module/cpyext/bufferobject.py: Fix for 0e27a73c87f6

16:05 aboudreault has quit [Excess Flood]

16:08 kipras`away is now known as kipras

16:09 aboudreault has joined #pypy

16:13 aboudreault has quit [Excess Flood]

16:16 aboudreault has joined #pypy

16:30 <bbot2> Failure: http://buildbot.pypy.org/builders/pypy-c-jit-linux-armhf-raspbian/builds/1323

16:41 <kenaan> arigo default ac573ca24687 /pypy/: Make sure sys.prefix is always defined, even if it contains the translation-time value. Fix obscure issues where th...

17:13 <cfbolz> plan_rich_: I am getting this kind of error with vmprof and a nightly pypy:

17:13 <cfbolz> https://www.irccloud.com/pastebin/HUTbgumd/

17:13 arigato has quit [Quit: Leaving]

17:33 Baliano has joined #pypy

17:36 Baliano has quit [Client Quit]

17:36 <kenaan> planter default bed189ed5900 /lib-python/2.7/zipfile.py: improve zipfile performance by not doing repeated string concatenation (committed by cfbolz)

17:54 aboudreault has quit [Excess Flood]

17:55 aboudreault has joined #pypy

18:15 aboudreault has quit [Excess Flood]

18:17 aboudreault has joined #pypy

18:35 aboudreault has quit [Excess Flood]

18:41 aboudreault has joined #pypy

18:47 marr has joined #pypy

18:54 <plan_rich_> cfbolz, hi, I have fixed that issue in 0.4.6.dev1, I triggered an official build (0.4.6) that includes the fix

18:56 <plan_rich_> or if you install the pre release version then it should now work

19:02 tormoz has joined #pypy

19:03 mattip has left #pypy ["bye"]

19:07 <cfbolz> plan_rich_: thanks!