On Jan 4, 2007, at 2:45 PM, Johan Nordlander wrote:
Hi Andy, and Happy New Year!
Happy New Year!
I'm having problems accessing darcs.haskell.org, otherwise you would have seen some recent pushes from my side. Do you know the status of that machine and when we should expect to see it up and running again?
It seems to be working fine now. Can you email me as soon as you see any problem, because explicitly galois maintain this machine, specifically to help the GHC crowd.
Speaking of administration, I believe it would be a good idea to relay all darcs push messages to the timber list (we can probably do without a specific commit list at this stage). Is this easy to fix from your side?
I'll look into this. Does anyone know how to do this?
Thanks! I'll look forward to more coming down the pipe.
Is there something you would like me to look at/add? I planning on adding some parser tests as a start, but there might be something better to do.
All kinds of tests would be valuable, we need to establish a test suite rather quickly. Building a test suite for the parser is probably also a good opportunity to get acquainted with the finer details of the language ("what on earth is this construct supposed to mean?"). Moreover, there is much room for improving the grammar, e.g. for the purpose of avoiding the shift/reduce conflicts we currently have (or at least understanding them better). So you're very welcome to work on that part. However, my pending patches contain some changes to the parser, so you might want to wait for my update.
I built a trivial test framework, and I'll check to see if it is checked in to the timber darcs. It should be easy to use to test the parser. I'm also working on a coverage tool for Happy (via the Hpc work), so we can check that we've actually use every production in our grammar in our tests.
Another good place to look at would be the Kindle2C pass, once I'm able to commit the Core2Kindle conversion (at least that might make the connection between KIndle and Timber clearer). It's not very far away now, but doing it properly has required some changes to the way the "self" variable is handled by the front end (and has even suggested some uninteresting changes to the language itself...!). I'll still need a couple of days to finish it off.
Then we must update the lambdalifter to the changes made to Kindle, but once that's done we should essentially have a running compiler! However, very little testing has been done, so we'll probably have to spend quite some time hunting bugs and building up a test suite. My intention is to ease that process by adding extensive comments to all modules, along the lines of my recent changes to Syntax2Core.
BTW, I suggest we move all these discussions to the timber list from now on. Some of our recent conversations might also be of interest to the others, so perhaps we should forward a few messages to the list if it's ok with you.
Sure. I'm happy to have any of these discussions on the timber list. (I've forward this email to timber@haskell.org. I'm looking forward to the working compiler. As we add tests, we should be able to make progress quickly. There are also rumblings of an Ajax based Haskell debugger that could be used by Timber without to much extra effort; the main issue is getting accurate span locations into the Timber-level AST. AndyG