mandeep_ has quit [Remote host closed the connection]
mandeep has joined #pypy
alawrence has joined #pypy
<alawrence>
ronan: Are you attending PyCon UK? Would it be a good idea to hold a PyPy sprint there?
alawrence has quit [Quit: Page closed]
<njs>
https://bitbucket.org/pypy/pypy/issues/2943/ is becoming more of an issue for us now that pypy3.5 is gone... basically we don't have any way to get accurate pypy coverage :-/
<njs>
(^^ cc nedbat)
<cfbolz>
njs: thanks for the ping, I should look into it more
<cfbolz>
njs: there is no smaller example anywhere yet though, is there?
<njs>
cfbolz: no sorry
<njs>
cfbolz: though when we saw similar things in 3.8, it was because of weird lnotab results
<cfbolz>
yes, I suspect it's something like that
<njs>
cfbolz: so it might be possible to just grab one of the files thats acting weird, compile to .pyc, and see if the lnotab is right
<cfbolz>
njs: probably a corner case in the wordcode support
<njs>
on the upper right there are some buttons labeled "pypy36nightly", "cpython", "pypy36release" that you can click to toggle including coverage from each in the view
<njs>
line 170 is a comment, and marked as uncovered iff either of the pypy buttons is enabled
<cfbolz>
njs: yes, it's pretty weird that it is even considered a coverable line
<njs>
cfbolz: yeah, that's the bug... I *think* this means that there is some offset in some code object that claims to map to that line
<cfbolz>
right
lritter has joined #pypy
* fijal
waves
mandeep has quit [Remote host closed the connection]
<cfbolz>
njs: sidenote: should the tests pass?
<cfbolz>
pff, and I can't reproduce the problem locally either :-(
azrdev has left #pypy ["WeeChat 2.4"]
antocuni has joined #pypy
<cfbolz>
ok, now I can
themsay has quit [Ping timeout: 268 seconds]
inhahe has quit [Ping timeout: 246 seconds]
xcm has quit [Remote host closed the connection]
antocuni has quit [Ping timeout: 245 seconds]
antocuni has joined #pypy
xcm has joined #pypy
<cfbolz>
yeah, something is definitely wrong with lnotab
<cfbolz>
I am noticing it while debugging the problem that the debugger thinks I am in the wrong line
themsay has joined #pypy
ambv has joined #pypy
xcm is now known as Guest68271
Guest68271 has quit [Killed (hitchcock.freenode.net (Nickname regained by services))]
xcm has joined #pypy
themsay has quit [Read error: Connection reset by peer]
themsay has joined #pypy
<arigato>
yay more lnotab issues
<cfbolz>
arigato: did you find some too, recently? or just in general?
<arigato>
not recently, I'm thinking about a very long time ago where I found and fixed a CPython issue
<cfbolz>
arigato: right
<cfbolz>
this one is clearly about pypy3
<cfbolz>
trying a nice test case reducer to get a minimal file that shows the problem
<cfbolz>
seems to work very nicely
<nedbat>
does this mean i am off the hook? :)
<cfbolz>
nedbat: unclear ;-)
<cfbolz>
so far it definitely looks like a pypy lnotab bug, yes
<cfbolz>
nedbat: I can definitely produce a bug in pypy3's dis module, so I'd say you're off the hook :-)
<nedbat>
cfbolz: :)
tsaka__ has joined #pypy
<kenaan>
fijal copystrcontents-in-rewrite 825a9d694c31 /rpython/jit/: an attempt to rewrite copystr/copyunicode content
mattip has quit [Ping timeout: 258 seconds]
mattip has joined #pypy
dddddd has joined #pypy
<kenaan>
mattip default 5604ceb9aa78 /pypy/objspace/std/unicodeobject.py: win32, python2 doesn't like surrogates, skip this check
<njs>
cfbolz: re: the tests passing, I have a theory that there's something wrong with the combination of trio's testsuite + pypy + openssl 1.1.1, but haven't followed up yet. (everything passes in CI but not on my laptop, and I think it's the openssl differences)