[plt-dev] documentation slowing down installs and builds dramatically

From: Robby Findler (robby at eecs.northwestern.edu)
Date: Mon Jun 1 09:58:46 EDT 2009

I think PLaneT does indeed work that way. Even worse, there is a bug
somewhere that causes CM to not compile package B (or C or D ...)
making things even worse. I got as far as figuring out that it was a
parameter that changed (I think it was the namespace?) that somehow
causes CM to disable itself, but I could not figure out what the right
fix was. I believe the right fix is to be using a new instantiation of
CM and the code seems to be doing that, but apparently it isn't.

(I should probably go try again to fix that. I got lost somewhere
inside setup-plt last time I tried.)

Robby

On Mon, Jun 1, 2009 at 8:50 AM, Matthew Flatt <mflatt at cs.utah.edu> wrote:
> I haven't investigated, but I have a guess. When a planet package A
> depends on another planet package B, then trying to compile A triggers
> an install of B. As the docs for B build, the search and master index
> are re-built, and that's where the time goes. Then the build of A
> continues, and again the search and master index are rebuilt. If you
> end up installing N packages, then the search and master index are
> built N times, which would take a while if N is big enough.
>
> Maybe someone who knows planet better can say that it doesn't work that
> way? Or maybe someone has observed enough about the build to say that
> the time isn't spent just in re-building the index and search?
>
> At Sun, 31 May 2009 00:14:59 -0400, Neil Van Dyke wrote:
>> I have a fresh anecdote of practical impact of the slow documentation
>> crunching.
>>
>> The first bit of code I try to run from the Snooze tutorial takes over 9
>> minutes on a fast computer, before it attempts to connect to the database.
>>
>> It installs a lot of small PLaneT packages in these 9+ minutes, and
>> seems to spend most of that time crunching documentation.
>>
>> (That's with reasonably fast Internet, PLTCOLLECTS unset, DrScheme
>> memory limit off, lots of RAM, and no swap space in use, otherwise idle
>> 2GHz dual-core, on Linux.)
>>
>> --
>> http://www.neilvandyke.org/
>> _________________________________________________
>>   For list-related administrative tasks:
>>   http://list.cs.brown.edu/mailman/listinfo/plt-dev
> _________________________________________________
>  For list-related administrative tasks:
>  http://list.cs.brown.edu/mailman/listinfo/plt-dev
>


Posted on the dev mailing list.