Minutes of the Mercury meeting Fri Nov 17th, 3:15pm to 4:10pm.

Attendees:
	Fergus Henderson 
	Zoltan Somogyi 
	David Jeffery 
	Mark Brown 
	David Overton 
	Simon Taylor 
	Kevin Glynn 

Apologies:
	Tyson Dowd  (overseas)

1.  Sourceforge:

	The discussion on what we should do with respect to putting Mercury
	on Sourceforge was left until the next meeting, when Tyson Dowd should
	be back from overseas.

1.  Release plans:

	We discussed a series of issues that have come up on mercury-developers
	concerning the upcoming release.

	- streams interface
		It was generally agreed that the streams interface that Peter
		Ross has been developing should go in the extras directory
		for this release. This will give users some chance to 
		experiment with it before we think about what should go in
		the standard library.
	- `enum' type class
		Simon Taylor has proposed an `enum' type class (which he has
		used as part of the sparse_bitset implementation). Some
		discussion ensued about whether the name and structure of this
		class is appropriate (given that it will be visible in the
		standard library). The issues raised were:
		- Given that the proposed class only has methods for conversion
		to and from integers (with `from' being a semidet operation),
		there is really no requirement that the integers mapped onto
		be contiguous. In that case `enum' is perhaps not a good name.
		- Perhaps we should have a class for `dense' enumerations, that
		has for example methods to indicate the upper and lower bounds
		that the enumeration maps onto.
		- Perhaps we should use the same classes as Haskell as some
		people may be used to that hierarchy. It was pointed out that
		the Haskell class hierarchy is likely to undergo quite some
		change on the way to Haskell 2.

		It was decided to just go with Simon's original proposal ---
		there is no real problem with using the `enum' name, and
		no-one could think of a better one.

	- sparse_bitset
		Has been implemented and will be included for the release.

	- type specialisation for the library
		Has been implemented and will be included for the release.

	- Tag switches for the MLDS back ends
		Has been implemented for the hlc grades but not for the IL
		backend. It will be included as is for the release.

	- Foreign code mechanism
		Tyson has posted a diff, and this should be included for the
		release.

	- .NET `tech preview'
		This relies on the foreign code mechanism to implement some
		parts of the library for the IL backend.

		This can probably go in the release, but Tyson wasn't there
		to comment.

	- extended record syntax (allowing arguments to field names)
		Simon will implement this before the release.

		There was some discussion of whether we should allow field
		names to be overloaded within a module. It was decided that
		this is too much work to get done before the release.

	- @ syntax 
		Will not be done for the release (still needs thought, and
		possibly isn't a good idea at all).

	- Mmake rules for the IL backend
		These will be done for the release.

		If possible, these rules will only be generated when the
		IL back end is actually being used. There was some discussion
		about whether it would be worth it, given that `mmake depend'
		would then need to examine grade options or other flags. Fergus
		pointed out that the hlc back end has special dependencies
		generated for it, so it shouldn't be too difficult.

	- Tabled I/O for debugging
		There has been quite a lot done on this, but it won't get
		worked on for at least another week (as Zoltan will be busy
		marking exams until then). This should make it into the
		release.

		Practical experience is needed with the feature before we can
		decide whether or not it is advisable to turn it on by
		default.

	- `program representation' in the declarative debugger
		Mark Brown reported that his change to use the program
		representation data structure in the declarative debugger
		would be at least partly complete in time for the release, 
		but there is unlikely to be any user features added by then

3.  Release process:

	In about two weeks (ie. in early December), a CVS branch will be
	created for the release version. The intention is that after this
	time, only bug fixes will be checked onto the branch.

	The release (codename: Rudolph) is to be made before Christmas.

	- testing
		Fergus pointed at that with the release on a separate branch, 
		we will need to change our nightly testing scripts to do at
		least some of the testing on the release branch.

		It was decided that some of the linux machines should run the
		release branch and some the main branch. The alpha and
		sparc can just run the release branch (until a little while
		after the release).

	- outstanding bugs that need to be fixed before the release
		- hlc doesn't work properly on the alpha. Fergus will
		  investigate.
		- there is a bug that has been discussed on mercury-developers
		  to do with cse and existential types. Zoltan and David 
		  Jeffery will investigate
		- there is a bug to do with existential typing recently 
		  reported by Peter Ross. David Jeffery has a partial solution,
		  and will investigate further with Fergus.
		- there is a long outstanding bug to do with type classes
		  whereby do_call_class_method inserts the type infos and
		  type class infos in the wrong order. David Jeffery to
		  investigate.

4.  Testing on extras and samples:

	Zoltan suggested that we need a standard way of testing the
	extras and samples. Fergus suggested that we enforce a rule that
	each Mmakefile contain a `check' target. It was pointed out that
	it can be quite difficult to test some of the things in extras
	such as the Tcl/Tk interface. Zoltan suggested that we just test them
	by compiling to C code.

5.  Summer students:
	
	The summer students are starting on November 27th, and we intend to
	give them a couple of days of talks about Mercury and the compiler
	(and our SE processes). Zoltan will not be available on Monday the 27th
	or Tuesday the 28th, so someone else will need to give the talks.
	Fergus should be around, as will Tyson (we expect). This needs to be
	sorted out in the next week.

	David Overton mentioned that there may not be enough keys to the
	Mercury office (and that the door that there are keys to doesn't
	seem to work properly). David and Zoltan will sort this out with the
	office.

	After giving the summer students the introductory talks for a couple
	of days, Zoltan suggested that we set them a programming exercise for
	the rest of their first week. The battleships task that was a 433-257
	project (and was posted on mercury-users a while back) was suggested.

	Zoltan needs to organise the theatre in which we will give the
	presentations on the Monday and Tuesday, and needs to get the slides
	to those who will present.

5.  Progress reports:
	We went around the table, with each person in turn discussing what
	they had been working on recently and/or what they were planning
	to work on (other than the matters already discussed with respect
	to features to be put into the release).

	David Jeffery :
		- Has been fixing various type class bugs, but mainly working 
		on HAL.

	Zoltan Somogyi :
		- Has been marking exam papers, and will be doing so for the
		next week or so.

	Fergus Henderson :
		- Various little things, tag switches for MLDS

	Kevin Glynn :
		- By his own admission doesn't do anything

	Simon Taylor 
		- Working on getting aditi to call Mercury

		- Has been working on speeding up the compiler in various
		ways (these have been posted to mercury-developers). He
		reported that the overall speedup when compiling make_hlds.m
		is around 25%! Everyone agreed to buy him a beer.

		Zoltan suggested that this may be worth mentioning in the
		NEWS file (depending on how it shapes up compared to version
		0.9.1).

	David Overton :
		- Has been marking SE projects. 

	Mark Brown :
		- Has been working on using the program representation in
		  the declarative debugger.


Minutes taken by David Jeffery .