Perl 6 - the future is here, just unevenly distributed

IRC log for #marpa, 2014-07-21

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

All times shown according to UTC.

Time Nick Message
00:48 * Aria laughs.
00:48 Aria It's true.
00:48 Aria Actually, maybe we owe Bjorne something after all.
00:48 Aria Could be C++'s purpose is to serve as a warning to others.
00:59 jeffreykegler On the other hand ...
01:00 jeffreykegler At Peter Stuifzand's suggestion, I've been slowly listening through Stephanov's "Programming Conversations" on youtube.
01:01 jeffreykegler Stephanov has a lot interesting to say, and C++ continues to evolve.
01:02 jeffreykegler Not often, but sometimes an overhyped technology evolves to live up to its promises.
01:03 jeffreykegler Java, for example was presented as an alternative to C, which in its original versions was a joke.
01:03 jeffreykegler But now a Java programmer only pays a 10% performance penalty on comparable "down to the metal" type code ...
01:05 jeffreykegler and for most tradeoffs, 10% is a very acceptable cost.
01:09 ronsavage +10% /speed/ may be acceptable, but what about the human cost to get to this point? Here I mean in learning effort, intermediary versions which were slower, library incompatibilities, ...
01:13 jeffreykegler ronsavage: All true, but if speed was on the list of your objections to Java, the list is now one item shorter.
01:15 jeffreykegler ... and we can thank them for a proof-of-concept of the JIT idea.
01:18 jeffreykegler But in any case it is amazing ... 40 years later, the C language is still setting the target for others to aim at.
01:20 Aria I'd rather writeYup.
01:21 Aria heh. Two sentences.
01:21 Aria I'd rather write C to get near the metal.
01:21 Aria Or, perhaps, rust.
01:21 jeffreykegler Aria: I think I may have just noticed the pun in rust's name.  Is it deliberate?
01:21 Aria I ... don't know.
01:22 jeffreykegler Because rust is certainly about as close to the metal as you are going to get.
01:22 Aria Yup.
01:22 Aria I like it so far. Just playing with it.
01:22 Aria But it's got promise for that space. I think not sweeping as much under the rug as Go
01:23 Aria rust could be the next C, if Go's the next Java.
01:44 jeffreykegler On another topic, I've uploaded a new Marpa::R3, which will be Kollos-based.
01:44 Aria Ooh.
01:44 jeffreykegler ronsavage: you probably don't want to bother testing the Marpa::R3's -- I'm breaking compatibility.
01:46 jeffreykegler Those of you who have worked on porting Marpa to Windows may want to start thinking about how to repair all the damage I did to Windows portability when I converted the build to EU::MM :-)
01:47 jeffreykegler The EU::MM based Marpa::R3 has not yet gotten any reports from CPANtesters -- apparently there is some sort of glitch ...
01:47 jeffreykegler either that or the new build logic has stunned the bots into silence. :-)
04:02 ronsavage Funnily enough, I can easily resist the urge to test Marpa::R3 at this point in time.....
04:03 jeffreykegler joined #marpa
04:06 jeffreykegler ronsavage: Since you've been the most diligent tester of Marpa::R2, I worried you might leap at this one. :-)
04:07 jeffreykegler I try to make it a rule: the more helpful someone is, the more respect I show for the value of their time.
04:20 jeffreykegler Re rust: what is the management model?  BDFL (with the right to fork) like Linux, Perl 6, Lua and Golang?  Or more devolved like Perl 5 and Moose(?).  Specifically is there one person who gets the last word about what goes into rust?
05:31 ronsavage joined #marpa
05:49 jeffreykegler joined #marpa
05:52 jeffreykegler First results are in from CPANtesters for the EU::MM build version of R3 -- all "unknown", which may indicate an issue.
05:53 jeffreykegler I tested on 3 platforms before release, but it has happened before that I've had 100% success on my tests, and 100% non-successes on CPANtesters.
05:55 jeffreykegler All the CPANtesters results so far are from a single source, so it is possible that the EU::MM R3 is OK, but it certainly does not look good.
05:56 jeffreykegler If perhaps some of you might try a download and install of the latest Marpa::R3, and let me know?  Thanks.
05:58 jeffreykegler (I'm still advising against testing R3 against your own modules.  Since I make no guarantees of compatibility, I don't expect those tests to succeed.  But R3 *should* pass its own test suite.)
07:35 ronsavage I got my sys admin friend to install VirtualBox on my old laptop, and sometimes I can get Mint Linux running. If and when (again) I could try R3, but I'll have to set things up first.
07:35 ronsavage The VM also runs MS Win 7, but.............................
10:10 jeffreykegler joined #marpa
10:10 jeffreykegler http://www.cpantesters.org/cpan/report/a44dfab8-1096-11e4-a6c1-5f16e1bfc7aa
10:12 jeffreykegler The logs from CPANtesters are now in.  The problem is that two sets of objects are being created one with PIC and one without, and that in creating the static library, the ones without PIC are being used.
10:15 jeffreykegler (PIC means "position independent code" which is necessary for code to go into a dynamic library.  The XS module is a dynamic library.  As an intermediate step, a static library is created, and these, *if* they are the ultimate destination do not need PIC.  But this static library is a stage in the creation of a dynamic one, so it *does* require PIC.)
10:16 jeffreykegler There's an option to always use PIC, even for static libraries.  My guess is that, in the conversion, I neglected to use it.
10:17 jeffreykegler I'll check this guess out at a more reasonable hour in the morning, California time.
10:20 jeffreykegler Why did this work on my test setups?  Many configurations are forgiving of PIC/non-PIC mismatches when linking, and will change non-PIC code to PIC at link time.  Perhaps all 3 of mine are of the forgiving kind, but none of Slaven Rezić's are, which would explain 100% success on my platforms and 100% non-success on his.
10:21 jeffreykegler Back to sleep!
18:31 jeffreykegler joined #marpa
18:52 jeffreykegler I've just uploaded R3 3.003_002, and the tests on Slaven Rezić are coming thru OK this time.
18:53 jeffreykegler No doubt there'll be a lot of nits to be fixed in the conversion to EU::MM, but this is a good start.
23:23 jeffreykegler joined #marpa

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