| Note: these documents may be out of date. Do not use as reference! |
To see what is currently happening visit http://www.perl6.org/
Welcome to the third of my US tour Perl 6 summaries. Once again I'm pleased to report that the denizens of the Perl 6 mailing lists continue to make the life of a touring summarizer an easy one by not posting all that much to the lists. So, I can sit here in my room at the Shaker Inn in Enfield and marvel at the traffic noise outside, wonder about the car next door with the New Hampshire plates reading PERLFAN, and just generally appreciate the loveliness of the room.
But, while I'm doing that, I should start with perl6-internals
At the end of last week, Dan outlined his thoughts on how exception handling will work in Parrot. This week, people talked about it. Discussion revolved around how much information should be attached to an exception and how/whether we should support resumable exceptions.
groups.google.com[63.120.19.221]
Last week I said that "I get the strong feeling that Leo Tötsch
isn't entirely happy with the new Continuation Passing Style". This
week Leo corrected me; I hadn't noticed that the speed issues had been
addressed by the latest changes to parrot (in fact the current CPS
implementation is faster than the old invoke/ret scheme.
Sean O'Rourke addressed Leo's problem with the Perl 6 Compiler tests failing by saying that the compiler should really be ported to use CPS rather than implementing a new variant of the Sub PMC that uses the old scheme. Leo reckoned that such a port wasn't currently doable because IMCC needed to be modified to use the CPS scheme, which would also involve reworking the register allocator. Given Leo's prodigious rate of implementation, this may have already happened.
Clinton A. Pierce had reported a memory leak in Parrot, but tracked it down to a situation where he was doing:
.arg 0
call _foo
And forgetting to take the 0 off the stack. However, even after he'd fixed that, he had segfault issues, and posted a (largish) code fragment that tweaked the bug.
It appears that Parrot wasn't throwing warnings when stacks get to big, just failing silently. Leo added a check for too deeply nested stacks, which at least avoids segfaulting on logic bugs.
Leo and Dan discussed other places where such limit checking should be put in place. Dan also muttered something about turning stack chunks into PMCs, allowing for the garbage collection of stack frames. Leo also muttered about the proliferation of stack implementations in Parrot (there are five) and thinks it should be possible to have one general stack engine.
+ a unary operatorBernhard Schmalhofer found a problem with the Perl 6 implementation.
print +42, "\n";
printed '42', but omitted the carriage return. He fixed this by making
+ into a unary operator as well as a binary operator and sent the
patch to the list, where it was applied. Good catch Bernhard.
Jürgen Bömmels is in the process of porting the IO
subsystem from its current mem_sys_alloc/free based implementation
to the sunny, garbage collected, uplands of a PMC based
implementation. However, he's run into a problem; some of the
operations in op.ops use integer File Descriptors, grabbing
information from a table in the interpreter structure. This gets in
the way of garbage collection, since any integer could be a file
descriptor.
Jürgen proposed removing the integer file descriptors and
mandating that ParrotIO PMCs be the only way to access IO (including
the standard STDIN, STDOUT and STDERR). He proposed adding
get_std[in|out|err] ops to get at the standard streams.
Dan suggested that Jürgen Just Do It; the current IO system being more than slightly hackish; essentially put in place until something better came along.
Want to get involved in the Parrot development process? Don't know much about Virtual Machine design and implementation? Do know Perl? Dan has a small but interesting task for you.
At present, Parrot gets built without any compiler level optimizations turned on because files like tsq.c can't have any optimizations turned on (tsq.c is the thread safe queue module, which is "annoyingly execution-order-dependent because it has to operate safely as interrupt code potentially interrupting itself").
Dan would like a version of Configure.pl which can build a makefile (or whatever build tool we end up using) with per-C-file compiler flags, and it needs to be possible to override those flags, per file, by the platform configuration module.
Interested? David Robins seems to be, and he asked whether the build system had to be makefile based. Dan says not, but the really important thing is that the resulting build script, or the config system that generates the script be adequately understandable/maintainable.
groups.google.com[63.120.19.221]
.local and IMCCBugfinder General Clinton A Pierce is getting a headache trying to
understand .local. When he executes the following code
.local int f
.sub _main
.local int x
.sub _foo1
f=1
x=2
call _foo2
end
.end
.sub _foo2
print "f is 1: "
print f
print "\n"
ret
.end
.end
the output looks like:
f is 1: 2
Which isn't quite what one would expect.
Leo explained what's going on; essentially it boils down to issues
with register allocation not being aware of .local scopes. He
recommended that Clint use either true globals or lexicals instead of
.local. Clint isn't so sure that this is a good idea, pointing out
that there are occasions when having lexically scoped names at the IMCC
level as well as at the level of lexical pads would be very
useful.
"In my mind, when I saw: 1. .local, 2. automagical register
spillage in IMCC, and 3. nested compilation units I thought I'd found
Assembler Manna." -- Clint Pierce
groups.google.com -- Clint is puzzled
groups.google.com -- Leo explains it all
Luke Palmer has been thinking about value and reference objects. He
wondered if there was any value in a valclone operator alongside
set and clone which would allow the target PMC to decide whether
to use set or clone semantics. He also offered a patch
implementing the operator if people thought it would be useful. Leo
Tötsch wasn't sure the new operator was necessary.
Klaas-jan Stol noted that he'd encountered problems with reference/value confusion when he'd been working on his Lua compiler, but he wondered if the problem couldn't be solved by having a general, language independent "argument" PMC class. (I'm not sure I understood what he meant by this so I'm hoping for an explanation with code fragments).
There is a story that UK prime minister Harold MacMillan was asked by a student what it was that concerned him most as Prime Minister. Mac replied "Events dear boy, events."
Leo Tötsch laid out his thoughts and ensuing questions about Exceptions, events and threads, and how they played together. There has been a small amount of discussion in response to this, but I think everyone's currently thinking hard about the issue...
Luke Palmer wondered if there would be a standard way of inspecting
the call stack (for debugging/caller/etc). (I think I'm going to
switch to using the phrase 'call chain' rather than call stack, as the
presence of continuations makes the call 'stack' look pretty
unstacklike...).
Leo and Dan both thought that this would be a high level language issue rather than a Parrot issue, though Dan did note that there might be useful things that Parrot could do to make such introspection easier/possible.
Leo Tötsch has been thinking about occasions when one might need
to monkey with the internals of an existing continuation (he was
thinking about the 'warnings' state...) and proposed several
solutions. Dan favoured his new opcode, updatecc and thought it
would be good to be able to broaden the scope of what one could update
in a continuation/context. This scared Leo somewhat, but Dan came up
with some examples of where it might prove to be useful.
Almost nothing happened, there were all of 15 messages.
Miko O'Sullivan engaged in some summer daydreaming by asking what everyone was looking forward to most from Perl 6. Miko himself is looking forward to more Pure Perl modules; if Perl 6 delivers on its performance promises then there are going to be more and more things where implementing directly in Perl will be fast enough, and Perl is so much easier to implement in than C...
Jonathan Scott Duff incurred Cozeny when he said that he's hoping that by this time next year we'll have an 85% complete Perl 6 that will be usable in production (by brave people). Simon Cozens noted that we already have such a beast and it's called Perl 5. For some reason this led to a new marketing slogan being proposed: Perl 6, the reconstituted cheeseburger of programming languages. Somehow I don't think that one's going to fly. (I just read this bit out to my wife and she says that she really doesn't like the thought of a flying reconstituted cheeseburger, so I think we'd best leave it at that.
Tcha! I announce the retirement of Leon Brocard from his post as Perl 6 Summary Running Joke and put the right to choose the next joke up for auction at YAPC. And what do you know, the winner of the auction nominates Leon Brocard as the new running joke. So, settle in for another year of desperate rationalizations for mentioning Leon in these summaries. Who knows, maybe Leon's Parrot related workrate will go up to such an extent that it'll be easy, but somehow I doubt it.
Thanks to, in chronological order, Adam Turoff and Lisa Wolfisch; Walt Mankowski, Mark Jason and Laurie Dominus; Dave Adler; Dan and Karen Sugalski; and Uri and Linda Guttman for being such fine hosts in Washington, Philadelphia, New York, Hartford and Boston respectively. Next time we do this, we will not be attempting to visit quite so many cities on the Eastern Seaboard in such a short time. At one point all we were seeing was Perl nerds and freeways in rapid succession.
As ever, if you've appreciated this summary, please consider one or more of the following options: