Perl 6 - the future is here, just unevenly distributed

IRC log for #marpa, 2016-06-11

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

All times shown according to UTC.

Time Nick Message
00:07 idiosyncrat_ joined #marpa
00:09 idiosyncrat_ [2016-06-10T23:27:32Z] [Chris Williams (BINGOS)] [fail] [JKEGL/Marpa-R3-4.001_011.tar.gz] [x86_64-linux-thread-multi] [perl-v5.22.0] [e79cd4e6-2f62-11e6-ab41-c893a58a4b8c] [2016-06-10T23:27:3
00:09 idiosyncrat_ So Marpa-R3-4.001_011.tar.gz does seem to be failing everywhere.
00:10 idiosyncrat_ In a few hours I'll be able to look at the reports in more detail.
01:17 ronsavage joined #marpa
01:27 ronsavage 4.001_011 has massive failure on my Debian box:
01:27 ronsavage Files=85, Tests=100, 13 wallclock secs ( 0.20 usr  0.10 sys + 12.11 cusr  0.67 csys = 13.08 CPU)
01:27 ronsavage Result: FAIL
01:27 ronsavage Failed 81/85 test programs. 47/100 subtests failed.
01:27 ronsavage Makefile:1045: recipe for target 'test_dynamic' failed
01:27 ronsavage make: *** [test_dynamic] Error 255
01:48 ilbot3 joined #marpa
01:48 Topic for #marpa is now Start here: http://savage.net.au/Marpa.html - Pastebin: http://scsys.co.uk:8002/marpa - Jeffrey's Marpa site: http://jeffreykegler.github.io/Marpa-web-site/ - IRC log: http://irclog.perlgeek.de/marpa/today
02:01 idiosyncrat_ I think what Linus Torvalds says in these situations is "I blame everyone but myself."
02:01 idiosyncrat_ :-)
02:02 idiosyncrat_ ronsavage: Could you put the traces on shadowpaste, just so we can be sure it's the same problem.
02:07 ronsavage Gist: https://gist.github.com/ronsavage​/450d9edaf5404a4aa6dbd9b65bd8aaba
02:13 idiosyncrat_ I don't see the "Makefile:1045: recipe for target 'test_dynamic' failed"
02:13 idiosyncrat_ Perhaps you rerun without cleaning the directory?
02:14 idiosyncrat_ Which is fine if the build works like it should, but this is clearly not the case here, and the 1st failure is the most interesting.
02:14 idiosyncrat_ s/rerun/reran/
04:50 rns joined #marpa
05:00 kaare_ joined #marpa
06:28 rns left #marpa
07:12 idiosyncrat_ Good night!
09:44 maybekoo2 joined #marpa
13:24 maybekoo2 joined #marpa
14:56 idiosyncrat_ joined #marpa
14:57 idiosyncrat_ Marpa-R3-4.001_011 testing is close to pessimal -- it fails (almost) everywhere except my development boxes, leaving me no way to try to duplicate the issue.
15:10 idiosyncrat_ rns: Thanks for http://irclog.perlgeek.de/m​arpa/2016-06-10#i_12642266 and http://irclog.perlgeek.de/m​arpa/2016-06-10#i_12642386
15:10 idiosyncrat_ Re http://irclog.perlgeek.de/m​arpa/2016-06-10#i_12642386
15:11 idiosyncrat_ I've enabled warnings in standard Makefile, and have fixed many of these.  If you'd care to rerun, I can go after any which are left.
15:12 idiosyncrat_ Re http://irclog.perlgeek.de/m​arpa/2016-06-10#i_12642266
15:12 idiosyncrat_ This is the nearest thing to a clue about our problem, I think.  I wish I could duplicate it here.
15:25 rns joined #marpa
15:25 rns idiosyncrat: re http://irclog.perlgeek.de/m​arpa/2016-06-11#i_12647981 -- sure, will let you know.
15:31 shadowpaste "rns" at 217.168.150.38 pasted "01-lua.t with Marpa::R2 meta_recce({ trace_values => 2}); -- for comparison" (84 lines) at http://fpaste.scsys.co.uk/522746
15:34 shadowpaste "rns" at 217.168.150.38 pasted "01-lua.t with Marpa::R2 meta_recce({ trace_values => 99}); -- for comparison" (50983 lines) at http://fpaste.scsys.co.uk/522747
15:36 shadowpaste "rns" at 217.168.150.38 pasted "warnings under msvc 2010 after fixing some under gcc" (12 lines) at http://fpaste.scsys.co.uk/522748
15:38 rns idiosyncrat: the above are, in reverse, warnings caught by msvc after rerun with freshly update repo
15:41 rns and traces of the grammar in 01-lua.t when used with Marpa::R2 -- the short trace one looks similar, unlike the long -- hope this helps, feel free to ask if can help with more.
16:32 idiosyncrat_ rns: commit f8845e0 attempts to fix the MSVC warnings -- it tests OK on my developement machine.
16:34 idiosyncrat_ rns: A next step, perhaps, is to sit down with two traces -- one successful and one failing, and spot where the traces first diverge.
16:34 idiosyncrat_ If you could take a crack at this, I'd greatly appreciate it.
16:42 idiosyncrat_ Since you may need to compare across differents architectures, unfortunately simply diff'ing is problaby not going to work.  Instead we'll need to eyeball for a "real" divergence, as opposed to one that's an artifact of the architecture.
16:56 shadowpaste "rns" at 217.168.150.38 pasted "warnings under msvc 2010 after commit f8845e0" (5 lines) at http://fpaste.scsys.co.uk/522750
16:56 rns idiosyncrat: re warnings -- most are fixed, see above.
17:00 rns re traces comparison -- will it make sense to start from short traces on windows (pass) vs. cygwin (fail)? I'd do a tab-separated file for you to look at?
17:13 rgrinberg joined #marpa
17:24 idiosyncrat_ Kind of a minimal example, yes.
17:25 rns ok here we are -- the https://www.diffchecker.com/98zyjqpd
17:25 rns 01-lua.t trace_values => 2, left is windows (pass), right is cygwin (fail)
17:27 idiosyncrat_ The rules seem to be numbered differently, which is interesting.
17:28 idiosyncrat_ I usually try for an architecture-independent order, sometimes even at the cost of some efficiency.
17:28 idiosyncrat_ diffchecker.com looks nice, btw
17:30 rns yes diffchecker's good.
17:30 idiosyncrat_ Also, even though you set trace_va;lues the same, there are lost of messages on the red (windows) side which are not on the right (green) side.
17:30 idiosyncrat_ That's fishy.
17:31 rns yesp -- for one, I noted that passing trace starts with
17:31 rns statements[Seq] -> statement
17:31 rns vs. failing trace starting with
17:31 rns rhs[Seq] -> rhs primary
17:33 idiosyncrat_ I'm inclined, unless something obvious points to the issue, to first fix the rule numbering to make it architecture independent.
17:35 rns looks like a good idea
17:39 rns I'm thinking about why your dev machine doesn't produce failures -- linux failures on cpantesters are with gcc version 5.3.1 20160121, cygwin has 5.3.0 ---
17:39 rns naive as it is, but yours is smth.like that?
17:44 idiosyncrat_ I am gcc (Debian 4.9.2-10) 4.9.2
17:46 idiosyncrat_ A new commit, with another pass at the MSVC warnings:  1985d42
17:46 rns thanks, will rerun and let you know
17:54 rns re gcc FWIW, http://www.cpantesters.org/cpan/report​/cbf83202-2f2c-11e6-8cc5-d1cef647b0e0 is on debian 3.16.7-ckt25-2 gccversion='4.9.2'
17:56 idiosyncrat_ I'm open to alternative ideas, but my current plan is to stabilize the rule numbering and ask you to retest and re-diff
17:58 idiosyncrat_ But actually ...
17:58 idiosyncrat_ now that I look at it, the rule numbering *is* stable, but the order of the diagnostics is off.
18:00 rns ok -- just thought updating gcc, if possible, is relatively easy and may lead to reproduction of the failure.
18:02 rns anyway you have better judgement -- so I'd gladly help in tests/reruns/rediffs.
18:03 idiosyncrat_ I wrote the bugs, so I have an advantage :-)
18:03 rns :)))) that's for sure, lol.
18:04 idiosyncrat_ As a working hypothesis, let's take that the problem is between the message in lines 1 and 2 of the diff.
18:04 idiosyncrat_ (It may be earlier, but I think perhaps ignore that for the moment.)
18:05 idiosyncrat_ Could you put some hi there's, to try to find out where the two diverge?
18:06 rns Meaning some warn()'s?
18:07 rns re http://irclog.perlgeek.de/m​arpa/2016-06-11#i_12648301 -- congrats, we have clean windows build now -- no warnings.
18:08 rns Not sure I understand that "put some hi there's" part, sorry.
18:18 kaare_ joined #marpa
18:31 kaare_ joined #marpa
18:42 kaare_ joined #marpa
18:44 idiosyncrat_ Yes, warn() or say STDERR or whatever you prefer
18:45 idiosyncrat_ That debuggin technique, of putting print diagnostics in the code, was old when I entered the business, and is called "using hi there's"
19:06 kaare_ joined #marpa
19:45 rns ok, got it, will do it and let you know.
20:06 idiosyncrat_ Thanks!
20:06 idiosyncrat_ Off on errands
20:06 idiosyncrat_ ALF
20:06 idiosyncrat_ s/ALF/AFK/ :-)
20:11 rgrinberg joined #marpa
21:25 rns idiosyncrat: tried to insert hi there's, but failed to produce anything meanigful -- or I'm looking at them wrong -- would you commit them so that I fetched and run them?
21:25 rns AFK till morning, will backlog.
21:25 rns left #marpa
21:31 rgrinberg joined #marpa
21:42 rns joined #marpa
21:51 rns idiosyncrat: filed a PR which fixes test failures under cygwin -- works for me under windows and cygwin -- not sure how you'd test it except by another dev version.
21:51 rns back to sleep, will backlog. :)
21:51 rns left #marpa

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