[plt-dev] Incredibly slow setup-plt

From: Carl Eastlund (cce at ccs.neu.edu)
Date: Fri Oct 9 01:22:19 EDT 2009

This worked.  A fresh build without the error ran 'make' in about 15
minutes and 'make install' in about 50.

Carl Eastlund

On Thu, Oct 8, 2009 at 9:19 PM, Matthew Flatt <mflatt at cs.utah.edu> wrote:
> `setup-plt' is not designed to work nicely when some core file is
> broken --- and the framework sounds core enough in this case. The setup
> process will keep trying to build things that rely on the broken file,
> and it will spend a lot of effort each time re-discovering the
> dependency and the broken file.
>
> I recommend that you just stop the build and update from SVN.
>
> At Thu, 8 Oct 2009 20:57:54 -0400, Carl Eastlund wrote:
>> I saw that it was fixed (after the build was started, of course).
>> After it finishes I may try another one in the background with the fix
>> in place to see if that had anything to do with the speed.
>>
>> And while I don't usually time my setup-plt, and it can certainly be
>> slow, I think 7 hours is a record.  Clean builds have taken 2 or 3
>> hours, but I've never had to wait essentially a whole work day for one
>> to finish before.
>>
>> Carl Eastlund
>>
>> On Thu, Oct 8, 2009 at 8:52 PM, Robby Findler
>> <robby at eecs.northwestern.edu> wrote:
>> > That file is fixed, at least. (Sorry no other guesses except to ask if
>> > this is new behavior or not.)
>> >
>> > Robby
>> >
>> > On Thu, Oct 8, 2009 at 7:46 PM, Carl Eastlund <cce at ccs.neu.edu> wrote:
>> >> I started a clean build (removed src/build, then ran configure, make
>> >> and make install) of the trunk today -- thankfully I've started
>> >> keeping two copies of the trunk, because the one I'm rebuilding is
>> >> still going, seven hours later.  Currently running Scribble files.
>> >> Now, I'm not on the world's fastest machine (2 GHz dual-core G5, 3G
>> >> RAM), but seven hours is over the top.  The only thing out of the
>> >> ordinary is an error I keep getting:
>> >>
>> >> /Users/cce/plt/two/collects/framework/private/keymap.ss:33:31:
>> >> compile: unbound identifier in module in: valid-key-bindings-lang?
>> >>
>> >> I suppose there are some files it keeps re-compiling until it hits
>> >> this error, but I wouldn't expect that to push it to seven hours.
>> >>
>> >> Anyone have an idea what might be causing this, or how I could diagnose it?
>> >>
>> >> Carl Eastlund


Posted on the dev mailing list.