ChanServ changed the topic of #picolisp to: PicoLisp language | Channel Log: https://irclog.whitequark.org/picolisp/ | Check also http://www.picolisp.com for more information
orivej has joined #picolisp
orivej has quit [Ping timeout: 246 seconds]
orivej has joined #picolisp
<tankf33der> Regenaxer: sounds good.
<Regenaxer> Hmm, not at all
<Regenaxer> No idea at the moment
<Regenaxer> Must manually check all related code again
<tankf33der> i will try decrease number of cpu and try run again, never seen crashes so far.
<Regenaxer> ok, good
<Regenaxer> I'm testing on an 8-CPU server
<Regenaxer> I see relatively many issues
<Regenaxer> Not crashes (segfaults) any more
<Regenaxer> but one process hanging, or the above IPC error ("upd" -- Undefined)
<Regenaxer> But it needs many runs
<Regenaxer> I did:
<Regenaxer> $ seq 9 |while read N; do echo $N; ./pil misc/stress.l -main -go -bye; rm db/test jnl db/test2; done
<Regenaxer>
<Regenaxer> i.e. 9 times
<Regenaxer> and that I run 5 times (thus 45 calls) until an error occurred
<Regenaxer> So really hard to reproduce
<Regenaxer> I released another version
<Regenaxer> Not really sure if it changes anything
<Regenaxer> Cleaned up some src variables in 'commit', it is not clear if there was a conflict
<Regenaxer> (not really sure what llvm does with the values ;)
<Regenaxer> I start the 9-tests again now
<Regenaxer> Perhaps it makes sense to increase the number of parallel processes?
<Regenaxer> Currently 60
<tankf33der> try to increase
<Regenaxer> I'll go to 99 if the current run passes
<Regenaxer> Takes a few hours
<Regenaxer> To be clear, a *single* run takes about 6 min
<Regenaxer> With 60 children
<Regenaxer> So a 9-test is about an hour, and multiple runs of these are necessary
<Regenaxer> I increase now to 99 children
<Regenaxer> But it may well be that less processes reproduce the error better. It is unpredictable
<Regenaxer> Crashes
<Regenaxer> 1
<Regenaxer> !? (wipe Lst)
<Regenaxer> ({2204}) -- Symbol expected
<Regenaxer> ....
<Regenaxer> 1087798 SIG-11
<Regenaxer> Random errors
<Regenaxer> Both the "upd" case and the above "Symbol expected" are obviously caused by IPC failure
<Regenaxer> Problem is just that it cannot be reliably reproduced, in a way that allows to pin it down
<tankf33der> writing fixedmath tutorial and found repl behaves strange on 2. input
<tankf33der> both pils.
<tankf33der> http://ix.io/2B7f
<Regenaxer> hehe, yes. *Scl is expected to be a number
<Regenaxer> The reader cannot check that all the time
<Regenaxer> (scl NIL) could be inhibited though
<Regenaxer> But not (off *Scl) or (setq *Scl) etc
<tankf33der> still thinking what way use to provide fixedmath info.
<Regenaxer> How to explain it?
<tankf33der> yea
<tankf33der> because form of explain important for me too
<Regenaxer> T
orivej has quit [Ping timeout: 240 seconds]
orivej has joined #picolisp
<tankf33der> Regenaxer:
<tankf33der> question:
<tankf33der> for (** 2 3.1) i need (pow) function from @lib/math.l, right ?
<Regenaxer> 'pow' wants 2 fixnums
<tankf33der> yea, right.
<Regenaxer> So it is fine
<tankf33der> ok
<Regenaxer> Not other way for a non-integer exponent
<Regenaxer> (no general way)
<tankf33der> yea. ok
<Regenaxer> :)
<Regenaxer> BTW, I'm just rewriting several parts of IPC (tell etc.)
<Regenaxer> Cleaning up
<Regenaxer> and I think I found also something not correct
<tankf33der> cool.
<Regenaxer> Let's hope ;)
<tankf33der> my tutorial is completed
<Regenaxer> Great!!
<Regenaxer> I'm testing now
<Regenaxer> Will you announce the tutorial in twitter and the list?
<tankf33der> sure, i will show you first :)
<Regenaxer> ok
<Regenaxer> I will release the modified IPC protocol now. There were some glitches in the byte stream concering embedded sizes. These happened when transferred in several chunks, which in turns happens if the other child process is busy and cannot accept the data
<Regenaxer> I did a local small test
<Regenaxer> Will now start my long 9-times tests repeatedly
<Regenaxer> to see if it is more stable now
<tankf33der> run test
<tankf33der> tests passed.
<Regenaxer> misc/stress still fail :(
<Regenaxer> One child just hangs
<Regenaxer> I did not know there is a Wikipedia article about fixpoint numbers :)
<Regenaxer> Cool tutorial!
<Regenaxer> Clever idea to call Ruby with (call "irb")
<Regenaxer> Perhaps some native English speaker can check the text?
<tankf33der> yea
<Regenaxer> I'm more testing stress.l, but it surely is not ok. Some deadlock it seems
<Regenaxer> Never-ending story :(
_whitelogger has joined #picolisp