2014-11-08T00:12:12Z Bicyclidine joined #sbcl
2014-11-08T00:48:44Z psilord1 joined #sbcl
2014-11-08T01:04:46Z kanru` quit (Ping timeout: 258 seconds)
2014-11-08T01:05:06Z Bicyclidine quit (Ping timeout: 244 seconds)
2014-11-08T01:11:10Z ehaliewicz joined #sbcl
2014-11-08T01:19:14Z specbot quit (Disconnected by services)
2014-11-08T01:19:18Z specbot joined #sbcl
2014-11-08T01:20:01Z jlarocco left #sbcl
2014-11-08T01:46:57Z krzysz00 quit (Ping timeout: 245 seconds)
2014-11-08T01:48:34Z ehaliewicz quit (Ping timeout: 250 seconds)
2014-11-08T01:54:08Z Hache_ quit (Remote host closed the connection)
2014-11-08T02:28:36Z krzysz00 joined #sbcl
2014-11-08T03:38:56Z christoph_debian quit (Ping timeout: 256 seconds)
2014-11-08T03:52:24Z psilord1 quit (Quit: Leaving.)
2014-11-08T03:52:24Z christoph_debian joined #sbcl
2014-11-08T03:56:58Z jlarocco joined #sbcl
2014-11-08T03:57:49Z jlarocco quit (Remote host closed the connection)
2014-11-08T03:58:13Z jlarocco joined #sbcl
2014-11-08T04:21:52Z loke_ joined #sbcl
2014-11-08T04:26:23Z Bicyclidine joined #sbcl
2014-11-08T04:32:43Z jlarocco quit (Quit: This computer has gone to sleep)
2014-11-08T04:47:45Z leo2007 quit (Remote host closed the connection)
2014-11-08T04:50:39Z jlarocco joined #sbcl
2014-11-08T04:53:25Z jlarocco quit (Client Quit)
2014-11-08T04:53:36Z jlarocco joined #sbcl
2014-11-08T04:54:04Z jlarocco quit (Client Quit)
2014-11-08T04:57:40Z scymtym_ joined #sbcl
2014-11-08T05:05:04Z karswell` joined #sbcl
2014-11-08T05:06:08Z karswell quit (Read error: Connection reset by peer)
2014-11-08T05:12:07Z krzysz00 quit (Ping timeout: 265 seconds)
2014-11-08T05:13:26Z krzysz00 joined #sbcl
2014-11-08T05:17:53Z Bicyclidine quit (Ping timeout: 264 seconds)
2014-11-08T05:18:22Z Bicyclidine joined #sbcl
2014-11-08T05:29:25Z zacts joined #sbcl
2014-11-08T05:38:01Z cmack``` quit (Ping timeout: 272 seconds)
2014-11-08T05:47:31Z Bicyclidine quit (Ping timeout: 272 seconds)
2014-11-08T06:01:40Z krzysz00 quit (Ping timeout: 244 seconds)
2014-11-08T06:03:17Z krzysz00 joined #sbcl
2014-11-08T06:06:10Z karswell` quit (Remote host closed the connection)
2014-11-08T06:09:59Z rpg quit (Quit: rpg)
2014-11-08T06:29:37Z oleo quit (Read error: Connection reset by peer)
2014-11-08T06:31:06Z oleo joined #sbcl
2014-11-08T06:49:27Z krzysz00 quit (Ping timeout: 245 seconds)
2014-11-08T06:49:40Z angavrilov joined #sbcl
2014-11-08T07:33:32Z krzysz00 joined #sbcl
2014-11-08T07:35:51Z psy_ quit (Remote host closed the connection)
2014-11-08T07:40:02Z psy_ joined #sbcl
2014-11-08T07:45:18Z leo2007 joined #sbcl
2014-11-08T08:00:17Z scymtym_ quit (Ping timeout: 245 seconds)
2014-11-08T08:00:51Z loke_ quit (Read error: Connection reset by peer)
2014-11-08T08:05:24Z loke_ joined #sbcl
2014-11-08T08:11:29Z gingerale joined #sbcl
2014-11-08T08:38:27Z stassats joined #sbcl
2014-11-08T10:34:47Z prxq joined #sbcl
2014-11-08T10:43:59Z DGASAU quit (Ping timeout: 245 seconds)
2014-11-08T10:45:47Z DGASAU joined #sbcl
2014-11-08T10:54:45Z edgar-rft quit (Quit: mental process closed because lifetime corrupted)
2014-11-08T11:06:04Z DGASAU quit (Ping timeout: 245 seconds)
2014-11-08T11:13:11Z DGASAU joined #sbcl
2014-11-08T11:23:52Z DGASAU quit (Ping timeout: 265 seconds)
2014-11-08T11:25:20Z DGASAU joined #sbcl
2014-11-08T11:56:52Z stassats quit (Ping timeout: 240 seconds)
2014-11-08T12:02:37Z edgar-rft joined #sbcl
2014-11-08T12:06:15Z Hache_ joined #sbcl
2014-11-08T13:00:19Z rpg joined #sbcl
2014-11-08T13:09:26Z kanru` joined #sbcl
2014-11-08T13:18:50Z krzysz00 quit (Ping timeout: 265 seconds)
2014-11-08T13:31:17Z scymtym_ joined #sbcl
2014-11-08T13:37:07Z leo2007 quit (Quit: rcirc on GNU Emacs 25.0.50.1)
2014-11-08T13:37:53Z kanru` quit (Remote host closed the connection)
2014-11-08T13:38:26Z kanru` joined #sbcl
2014-11-08T13:38:37Z leo2007 joined #sbcl
2014-11-08T13:43:36Z cmack``` joined #sbcl
2014-11-08T13:52:19Z stassats joined #sbcl
2014-11-08T14:19:53Z stassats: oops, i made char-downcase to be faster than upper-case-p
2014-11-08T14:29:12Z Krystof: and I am currently failing to find an unportability in genesis
2014-11-08T14:31:09Z stassats: argh, #\LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE
2014-11-08T14:31:29Z stassats: the only char that doesn't downcase to a single char
2014-11-08T14:32:54Z stassats: i̇, really?
2014-11-08T14:35:30Z stassats: that slows down my char-downcase up to upper-case-p
2014-11-08T14:36:27Z stassats: maybe have two *character-cases*, one for sb-unicode?
2014-11-08T14:40:51Z jlarocco joined #sbcl
2014-11-08T14:43:49Z stassats: a better idea, have a normal *character-cases* be a flat array, specialized to sb32, but twice as large, so that two positions would be used instead of conses
2014-11-08T14:44:18Z stassats: that occupies less space than *character-cases* with conses, and should be faster
2014-11-08T14:49:49Z stassats: and for sb-unicode **character-cases** it can store characters, not codes, since they are be code-chared anyway
2014-11-08T15:00:27Z stassats: was unicode designed by non-programmers?
2014-11-08T15:05:04Z attila_lendvai joined #sbcl
2014-11-08T15:15:16Z attila_lendvai quit (Quit: Leaving.)
2014-11-08T15:21:42Z attila_lendvai joined #sbcl
2014-11-08T15:24:00Z pkhuong: stassats: I think natural languages were designed by non programmers ;)
2014-11-08T15:26:32Z stassats: why is there #\MICRO_SIGN, which is not equal to #\GREEK_CAPITAL_LETTER_MU, but is sb-unicode:unicode-equal
2014-11-08T15:27:51Z Krystof: *blink*
2014-11-08T15:27:59Z Krystof: well that's a host leak that I wasn't expecting
2014-11-08T15:28:26Z Krystof: and I still don't know quite how it happens.  But, some time spent reading od output on core files later...
2014-11-08T15:28:49Z Krystof: sbcl-from-sbcl> (loop for x being the external-symbols of "CL" count 1) ; => 978
2014-11-08T15:29:07Z Krystof: sbcl-from-clisp> (loop for x being the external-symbols of "CL" count 1) ; => 1032
2014-11-08T15:29:19Z stassats: that's a tad too many
2014-11-08T15:31:17Z Krystof: oh but it's worse than that
2014-11-08T15:31:38Z Krystof: sbcl-from-clisp> (length (delete-duplicates (loop for x being the external-symbols of "CL" collect x))) ; => 978
2014-11-08T15:34:26Z attila_lendvai quit (Quit: Leaving.)
2014-11-08T15:34:31Z Krystof: looks like our handling of the CLOS package isn't quite right
2014-11-08T15:35:03Z stassats: so, when i define (defun 𝜇 ()), and then try to see its arglist or do M-., no dice
2014-11-08T15:35:43Z stassats: becuase (defun 𝜇 ()) => Μ, but (both-case-p #\𝜇) => NIL
2014-11-08T15:35:46Z stassats: fun!
2014-11-08T15:44:15Z stassats: and i suppose (sb-unicode:uppercase (sb-unicode:lowercase x)) isn't supposed to return x
2014-11-08T15:56:01Z stassats: and about the italic mu, it's not unly that it cannot be cased, but it's also normalized first into a greak mu, then upcased
2014-11-08T16:10:27Z krzysz00 joined #sbcl
2014-11-08T16:20:06Z jlarocco quit (Quit: This computer has gone to sleep)
2014-11-08T16:21:14Z stassats: ok, my flat array strategy is about twice as fast
2014-11-08T16:21:42Z stassats: compared to calling char-upcase-p and using the unicode infest case table
2014-11-08T16:30:02Z jlarocco joined #sbcl
2014-11-08T16:45:28Z zacts quit (Quit: leaving)
2014-11-08T17:02:28Z cmack``` quit (Ping timeout: 250 seconds)
2014-11-08T17:03:38Z jlarocco quit (Quit: This computer has gone to sleep)
2014-11-08T17:15:01Z jlarocco joined #sbcl
2014-11-08T17:17:44Z psilord quit (Quit: Leaving.)
2014-11-08T17:19:41Z kanru` quit (Ping timeout: 255 seconds)
2014-11-08T17:29:37Z karswell` joined #sbcl
2014-11-08T17:30:35Z jlarocco quit (Quit: This computer has gone to sleep)
2014-11-08T17:35:56Z slyrus joined #sbcl
2014-11-08T17:56:40Z karswell` is now known as karswell
2014-11-08T17:56:57Z heddwch quit (Ping timeout: 245 seconds)
2014-11-08T17:58:46Z heddwch joined #sbcl
2014-11-08T18:33:40Z cmack``` joined #sbcl
2014-11-08T18:35:30Z edgar-rft quit (Quit: memory access closed into eternal confusion)
2014-11-08T18:38:37Z Krystof: and done
2014-11-08T18:38:47Z Krystof: csr21@aleph-null:/tmp/tmp.YoTKDTBXf0$ cmp -i 1024 {sbcl,ccl32}/output/cold-sbcl.core
2014-11-08T18:38:47Z Krystof: csr21@aleph-null:/tmp/tmp.YoTKDTBXf0$ cmp -i 1024 {sbcl,ccl64}/output/cold-sbcl.core
2014-11-08T18:38:48Z Krystof: csr21@aleph-null:/tmp/tmp.YoTKDTBXf0$ cmp -i 1024 {sbcl,clisp}/output/cold-sbcl.core
2014-11-08T18:38:52Z cmack``` quit (Ping timeout: 265 seconds)
2014-11-08T18:50:47Z pkhuong: so... we could now use this as part of our regression tests? (:
2014-11-08T18:52:19Z Krystof: yes
2014-11-08T18:52:29Z Krystof: I believe that scymtym has the basic infrastructure for it set up
2014-11-08T19:02:23Z heddwch quit (Ping timeout: 240 seconds)
2014-11-08T19:03:52Z heddwch joined #sbcl
2014-11-08T19:09:09Z scymtym_: Krystof: with what i currently have, it would have to be somewhat outside SBCL's test suite since builds with different host compilers are performed independently of each other
2014-11-08T19:10:31Z scymtym_: my assumption was that a separate job would be triggered after all variants of a specific SBCL commit have been built, obtain the respective outputs and make the necessary comparisons
2014-11-08T19:12:02Z scymtym_: performing builds with multiple hosts in a single SBCL source tree and doing the comparison there would probably also work
2014-11-08T19:17:09Z Krystof: right, it's a sort-of meta regression test
2014-11-08T19:17:20Z Krystof: I can certainly integrate it into my release script
2014-11-08T19:17:59Z scymtym_: do you already have any scripts or recipes i could use?
2014-11-08T19:21:15Z Krystof: http://paste.lisp.org/display/144312
2014-11-08T19:22:57Z scymtym_: thanks
2014-11-08T19:23:48Z stassats: something is happening to pcl with my char-case changes, and of course the condition system isn't up for reporting
2014-11-08T19:25:21Z scymtym_: Krystof: and after that i just ensure that all 32-bit cores are identical and all 64-bit cores are identical?
2014-11-08T19:26:19Z Krystof: I'd say identical after the first 1024 bytes
2014-11-08T19:26:32Z Krystof: cmp -i 1024 a/output/cold-sbcl.core b/output/cold-sbcl.core is what I used
2014-11-08T19:27:42Z scymtym_: thanks, i will try that
2014-11-08T19:29:41Z oleo__ joined #sbcl
2014-11-08T19:31:07Z oleo is now known as Guest78785
2014-11-08T19:31:53Z Guest78785 quit (Ping timeout: 240 seconds)
2014-11-08T19:48:11Z scymtym_: stassats: in a long-running, cons-heavy and multi-threaded SBCL 1.2.4 process on x86_64, i saw an error like "invalid object is not of type system area pointer" in one thread and after that other threads were unable to gc. does that sound like the gc issue you fixed?
2014-11-08T19:59:32Z stassats: it does
2014-11-08T19:59:56Z scymtym_: that's a relieve
2014-11-08T20:00:16Z stassats: do you have something like ignore-errors?
2014-11-08T20:00:30Z stassats: or some kind of handler bind?
2014-11-08T20:00:57Z LiamH joined #sbcl
2014-11-08T20:03:02Z scymtym_: i couldn't look in detailed, but i suspect the error was signaled in one thread and handled in a HANDLER-BIND by trying to CONTINUE. after that a different thread seems to have tried to gc and run out of dynamic space during that.
2014-11-08T20:03:54Z stassats: so, a lot of consing and internal errors is exactly the recipe for that bug
2014-11-08T20:03:56Z scymtym_: i have a backtrace for that second thread but the original error is from memory (the process ran in a strange terminal)
2014-11-08T20:06:56Z scymtym_: i will try upgrading to master and see whether it happens again
2014-11-08T20:07:39Z scymtym_: seems very rare, though. the previous instance happened after running for more than 12 hours
2014-11-08T20:08:02Z stassats: 12 hours doesn't sound rare
2014-11-08T20:08:52Z scymtym_: under load though. allocation and gc should be exercised all the time
2014-11-08T20:09:23Z stassats: it also needs internal errors
2014-11-08T20:09:43Z scymtym_: since the component should be running 24/7 in the future, testing over longer periods of time will not be an issue
2014-11-08T20:10:16Z stassats: or interrupts
2014-11-08T20:10:27Z stassats: i suppose things like sigchld would exercise it
2014-11-08T20:10:32Z scymtym_: it has timers
2014-11-08T20:10:41Z stassats: non safepoint?
2014-11-08T20:10:58Z scymtym_: yes, just --fancy, i think
2014-11-08T20:11:09Z stassats: timers will do
2014-11-08T20:12:00Z scymtym_: i will try with master, thanks
2014-11-08T20:13:16Z stassats: alloc_number is used for emulating fp or what?
2014-11-08T20:13:26Z stassats: on non-ppc riscs
2014-11-08T20:15:21Z stassats: sigemt_handler in sparc-arch.c is strange
2014-11-08T20:15:29Z stassats: why does it print something to stderr?
2014-11-08T20:16:47Z scymtym_: this seems new: https://ci.cor-lab.org/job/sbcl-master/729/featureset=7,label=ubuntu_trusty_32bit/consoleFull (search for "fixup vector 0xcf97000 has a bad widetag: 38")
2014-11-08T20:16:56Z scymtym_: only seen on x86 so far
2014-11-08T20:18:13Z stassats: it does look unfamiliar
2014-11-08T20:19:38Z stassats: and it can only be on x86
2014-11-08T20:21:22Z stassats: and 38 is code-header
2014-11-08T20:21:50Z scymtym_: seems to happen in about one of two builds, but not totally deterministically
2014-11-08T20:22:15Z stassats: i changed that part in june
2014-11-08T20:23:03Z stassats: there is a magic constant there
2014-11-08T20:23:40Z scymtym_: i can try to bisect on my laptop later, the automated builds from that time are long gone
2014-11-08T20:25:08Z stassats: i think that's it
2014-11-08T20:29:35Z stassats: but i want a test case anyway
2014-11-08T20:31:58Z stassats: hah, slime gets confused by unbound markers
2014-11-08T20:33:32Z stassats: got some ldb action
2014-11-08T20:36:27Z stassats: it clearly is wrong but why isn't it triggered more often?
2014-11-08T20:40:15Z scymtym_: maybe it is deterministic but depends on the configured features
2014-11-08T20:42:31Z stassats: it's more of "it shouldn't work at all"
2014-11-08T20:53:41Z oleo__ is now known as oleo
2014-11-08T21:05:04Z stassats: ok, surprise, that constants_start_addr is never used
2014-11-08T21:07:42Z angavrilov quit (Remote host closed the connection)
2014-11-08T21:21:58Z stassats: bah
2014-11-08T21:22:07Z stassats: now it's my fault
2014-11-08T21:22:33Z stassats: wait, too quick for blame
2014-11-08T21:35:04Z stassats: ok, now that this magic constant appears to be unused, i don't see what may be wrong
2014-11-08T21:52:43Z oleo quit (Quit: Verlassend)
2014-11-08T21:56:10Z oleo joined #sbcl
2014-11-08T22:08:39Z stassats quit (Quit: ERC Version 5.3 (IRC client for Emacs))
2014-11-08T23:30:38Z prxq quit (Remote host closed the connection)