Re: [jamboost] how to skip regenerating the dependency tree

From: Rene Rivera (grafik.list_at_[hidden])
Date: 2004-04-28 12:21:30

Toon Knapen wrote:
> (I'm sorry, I know this has been discussed before but have found no link 
> to an explanation on how to use it).
> 
> I guess bjam first generates the whole dependency tree (parsing all 
> includes and following them) and then verifies the 'last change date' of 
> all nodes in this tree.

Yes.

> Now bjam takes quite some time before issueing its first compile or link 
> command so I figured: is there anyway to skip the recreation of the 
> dependency tree ?

Yes it takes a long time. And yes there is a way to cache the dependency 
scanning.

> Basically this comes down to equivalent of doing 'make' without first 
> running 'make depend'.

Good news is that bjam should already be doing this for you. Look for a 
file ".jamdeps", that holds a cache of all the depency scanning.

The bad news is that the slow startup has nothing to do with dependency 
scanning. But has everything to do with parsing and generating the 
actual target descriptions. It's a know problem for which we don't have 
a real solution. Bjam spends much time building many strings, as all the 
work it does is based on string and lists of strings. Boost.Build uses 
*many* strings and does *many* operations on them. Which results in a 
large memory use because bjam considers strings read-only, and doesn't 
ever collect unused strings (for various C reasons).

And we don't currently have a way to improve the situation. Ideas are 
welcome ;-)

-- 
-- Grafik - Don't Assume Anything
-- Redshift Software, Inc. - http://redshift-software.com
-- rrivera/acm.org - grafik/redshift-software.com - 102708583/icq