<jhass>
and sometimes there's also a null pointer access in run_poll
<jhass>
so nothing easy to reproduce
<jhass>
I suspect it's that the GC removes stuff passed to GLib as callback or/and GLib corrupting something when calling the callback
waj has quit [Quit: Leaving.]
waj has joined #crystal-lang
leafybasil has quit [Remote host closed the connection]
<asterite>
jhass: Ah, yes, that might be it
<asterite>
Maybe there would need to be a way to mark something as a GC root...
<jhass>
otoh. if it's a spurious GC, I would expect it to happen far sooner
<jhass>
it takes a few dozen iterations though
<jhass>
(the scheme is add callback, in the callback add the callback itself to be run in after the poll interval (currently 60 seconds as returned by GHs API)
<crystal-gh>
[crystal] ysbaddaden opened pull request #445: Added specific Errno exceptions (Errno::ENOENT, ...) (master...std-errno-classes) http://git.io/A6Vq
<jhass>
so it's confusing me a bit
<asterite>
So the callback is passed but crystal doesn't retain a reference to it?
<jhass>
it does, at least for a while
<jhass>
I shall compile with --debug and see if that spits out line numbers yet
<asterite>
No, debug doesn't add line numbers
<jhass>
:(
<jhass>
what does it even do then?
<waj>
jhass: if you set the env var GC_DONT_GC=1 you can disable the GC completely
<waj>
and figure out if that's the problem
<jhass>
thanks
<waj>
regarding line numbers… there is a tool in Mac that you can use to find line numbers from addresses: atos
<waj>
there might be something similar for linux
<jhass>
and that doesn't read it from the debug info but does what?
<waj>
the debug info contains the line numbers, but right now that info is not used when raising exceptions
<jhass>
looks like the atos equivalent is addr2line
weskinner_work has left #crystal-lang [#crystal-lang]
weskinner_work has joined #crystal-lang
asterite has quit [Quit: Leaving.]
bcardiff has quit [Quit: Leaving.]
<crystal-gh>
[crystal] asterite pushed 4 new commits to master: http://git.io/AiKb
<crystal-gh>
crystal/master 5720d19 Ary Borenszweig: Fixed some errors related to untyped fun literals and pointers
<crystal-gh>
crystal/master ee0811d Ary Borenszweig: Fixed: when inheriting a generic class you couldn't invoke its class methods
<crystal-gh>
crystal/master 7dbd2dc Ary Borenszweig: Fixed #442: incorrect "splatting a union (({Int32, Int32} | Nil)) is not yet supported"
ismaelga has quit [Remote host closed the connection]