Perl 6 - the future is here, just unevenly distributed

IRC log for #moarvm, 2016-10-30

| Channels | #moarvm index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
01:27 flaviusb joined #moarvm
01:51 FROGGS joined #moarvm
09:53 domidumont joined #moarvm
10:22 dalek MoarVM: f5d97c3 | FROGGS++ | Configure.pl:
10:22 dalek MoarVM: dont build dyncall if --has-libffi is supplied, fixes #436
10:22 dalek MoarVM: review: https://github.com/MoarVM/MoarVM/commit/f5d97c303d
10:49 kjs_ joined #moarvm
11:59 kjs_ joined #moarvm
12:29 FROGGS ./libmoar.so: undefined reference to `AO_fetch_compare_and_swap_full'
12:29 FROGGS collect2: error: ld returned 1 exit status
12:29 FROGGS on s390x...
12:32 jnthn FROGGS: My memory is hazy, but I think libatomicops is typically just macros on most platforms (and so maybe not linked in), but maybe it's defined as a function on s390x? Dunno though.
12:33 jnthn https://github.com/ivmai/libatomic_ops/blob/master/src/atomic_ops/sysdeps/gcc/s390.h#L44
12:34 jnthn Seems that actually it may be really missing
12:36 FROGGS jnthn: I'm using the lobatomic-ops debian package btw
13:21 FROGGS I think that's part of the problem: https://github.com/ivmai/libatomic_ops/blob/master/src/atomic_ops.h#L366-L377
13:21 FROGGS AO_HAVE_compare_and_swap_full *is* defined, just not one above, which we are using
13:22 FROGGS err, the one below is the one we are using
13:24 * FROGGS recompiles after applying a patch
13:42 FROGGS ohh
13:42 FROGGS src/moar.h:207:5: warning: implicit declaration of function 'AO_fetch_compare_and_swap_emulation' [-Wimplicit-function-declaration]
13:42 FROGGS AO_fetch_compare_and_swap_emulation(addr, old, newval)
13:42 FROGGS ^
13:42 FROGGS src/moar.h:207:5: note: in definition of macro 'AO_fetch_compare_and_swap_full'
13:42 FROGGS AO_fetch_compare_and_swap_emulation(addr, old, newval)
13:42 FROGGS that means it doesnt load the library right...
13:47 FROGGS LDLIBS    = -lffi -ltommath -latomic_ops -luv -lm -lpthread -lrt -ldl
13:48 FROGGS # ls -l /usr/lib/s390x-linux-gnu/libatomic_ops.a
13:48 FROGGS -rw-r--r-- 1 root root 4368 Oct 25 13:26 /usr/lib/s390x-linux-gnu/libatomic_ops.a
13:48 FROGGS I was under the impression that -latomic_ops would pick up this static lib?
13:55 domidumont joined #moarvm
15:08 zakharyas joined #moarvm
15:31 diakopter FROGGS: well is there a guard against loading the header twice?
16:05 FROGGS diakopter: no...
16:06 FROGGS diakopter: and after nqp and rakudo is build I try to protect it from loading twice...
17:17 FROGGS_ joined #moarvm
17:50 domidumont joined #moarvm
17:55 domidumont joined #moarvm
18:17 leego joined #moarvm
18:29 kjs_ joined #moarvm
18:51 FROGGS_ ewww, a lot of nativecall tests are failing on s390x :o(
18:52 lizmat endian issues ?
18:55 FROGGS_ don't know yet
18:55 FROGGS_ but we have other big endian machines where make test passes
20:10 kjs_ joined #moarvm
20:12 zakharyas joined #moarvm
20:22 MasterDuke joined #moarvm
20:38 MasterDuke in my playing around with backtraces, i added some printfs to MVM_exception_backtrace, https://github.com/MoarVM/MoarVM/blob/master/src/core/exceptions.c#L362. i'm printing out filename_str and line_number (and also the "value" variable with each)
20:39 MasterDuke however, filename_str and its "value" variable are always empty, while line_number has a value (that i can see in the Perl 6 backtrack output) but its associated "value" variable is also always empty
20:42 zakharyas1 joined #moarvm
20:47 dalek MoarVM: 48c2482 | FROGGS++ | src/core/nativecall_libffi.c:
20:47 dalek MoarVM: handle libffi return type more corrent
20:47 dalek MoarVM:
20:47 dalek MoarVM: Again, see https://github.com/libffi/libffi/issues/286
20:47 dalek MoarVM: review: https://github.com/MoarVM/MoarVM/commit/48c24822d0
20:49 FROGGS_ correct*
20:49 FROGGS_ damn it
20:52 japhb FROGGS_: 'correctly', I think
20:53 MasterDuke i also added some printfs to MVM_exception_backtrace_line, but they weren't showing up at all, but those i do see if i add --ll-exception
20:53 MasterDuke and those are populated with the values i see in the Perl 6 backtrace
20:54 FROGGS_ MasterDuke: I think I've observed the same in the past... but I cannot help you right now
20:56 FROGGS_ MasterDuke: so you're thinking that you printf wrong in MVM_exception_backtrace?
20:56 MasterDuke no worries
20:56 MasterDuke printf("MVM_exception_backtrace: filename_str = %s, value = %s\n", filename_str, value);
20:57 MasterDuke hmm, filename_str is a MVMString, but value is an MVMObject, might need to do something with it
20:58 FROGGS_ MVMString      *filename_str;
20:58 FROGGS_ you cannot print that
20:58 FROGGS_ MVM_string_utf8_encode_C_string(tc, filename_str)
20:59 MasterDuke ah, thanks
21:00 FROGGS_ ohh, and value just seems to be a temporary var that holds the same
21:00 MasterDuke FROGGS_++, now there's stuff printed
21:01 FROGGS_ \o/
21:15 kjs_ joined #moarvm
21:43 leego joined #moarvm
21:57 FROGGS_ phew, s390x is fine now...
21:59 leego joined #moarvm
22:00 FROGGS_ ohh, wait, have to push that libatomic_ops patch... well, will do that tomorrow
22:18 leego joined #moarvm
22:20 leego joined #moarvm
22:40 leego joined #moarvm

| Channels | #moarvm index | Today | | Search | Google Search | Plain-Text | summary